# Research Figures and Special Tables
197 independent papers and their locally held appendices/versions were checked. 867 figures and 123 special tables form 990 logical visual records, delivered as 1045 PNG files. Report figures are excluded.
[Searchable purpose gallery](<00_Gallery.html>) · [Source-quality PDF album](<00_Source_Quality_Figures.pdf>)
Existing image filenames, record IDs and the first 717 PDF-album pages are preserved. New native-source pages are appended.
## Coverage by source document
| Source key | Paper | Role/version | Source pages | Figures | Special tables |
|---|---|---|---|---|---|
| P057 | P057 | main | 12 | 1 | 0 |
| P058 | P058 | main | 9 | 5 | 0 |
| P059 | P059 | main | 22 | 3 | 0 |
| P060 | P060 | main | 14 | 3 | 0 |
| P061 | P061 | main | 17 | 3 | 0 |
| P062 | P062 | main | 10 | 3 | 0 |
| P063 | P063 | main | 9 | 0 | 0 |
| P064 | P064 | main | 10 | 2 | 0 |
| P065 | P065 | main | 10 | 4 | 0 |
| P066 | P066 | main | 11 | 3 | 0 |
| P067 | P067 | main | 8 | 2 | 0 |
| P068 | P068 | main | 38 | 7 | 2 |
| P069 | P069 | main | 18 | 5 | 0 |
| P070 | P070 | main | 12 | 9 | 0 |
| P071 | P071 | main | 9 | 0 | 1 |
| P072 | P072 | main | 10 | 1 | 0 |
| P073 | P073 | main | 16 | 5 | 1 |
| P074 | P074 | main | 17 | 7 | 0 |
| P075 | P075 | main | 10 | 0 | 0 |
| P076 | P076 | main | 11 | 1 | 0 |
| P077 | P077 | main | 13 | 2 | 0 |
| P078 | P078 | main | 14 | 7 | 0 |
| P079 | P079 | main | 14 | 1 | 0 |
| P080 | P080 | main | 27 | 6 | 0 |
| P081 | P081 | main | 22 | 3 | 1 |
| P082 | P082 | main | 16 | 5 | 1 |
| P083 | P083 | main | 20 | 10 | 0 |
| P084 | P084 | main | 21 | 2 | 0 |
| P085 | P085 | main | 18 | 7 | 0 |
| P086 | P086 | main | 9 | 1 | 0 |
| P087 | P087 | main | 8 | 0 | 0 |
| P088 | P088 | main | 13 | 4 | 0 |
| P089 | P089 | main | 14 | 1 | 0 |
| P090 | P090 | main | 10 | 5 | 1 |
| P091 | P091 | main | 9 | 2 | 0 |
| P092 | P092 | main | 10 | 2 | 0 |
| P093 | P093 | main | 16 | 4 | 1 |
| P094 | P094 | main | 8 | 0 | 5 |
| P095 | P095 | main | 17 | 4 | 1 |
| P096 | P096 | main | 13 | 3 | 0 |
| P097 | P097 | main | 12 | 2 | 1 |
| P098 | P098 | main | 8 | 0 | 0 |
| P099 | P099 | main | 18 | 5 | 0 |
| P100 | P100 | main | 13 | 2 | 0 |
| P101 | P101 | main | 11 | 1 | 1 |
| P102 | P102 | main | 13 | 2 | 0 |
| P103 | P103 | main | 15 | 2 | 0 |
| P104 | P104 | main | 12 | 3 | 0 |
| P105 | P105 | main | 9 | 2 | 0 |
| P106 | P106 | main | 40 | 0 | 0 |
| P107 | P107 | main | 13 | 1 | 0 |
| P108 | P108 | main | 14 | 5 | 0 |
| P109 | P109 | main | 9 | 1 | 0 |
| P110 | P110 | main | 10 | 3 | 0 |
| P111 | P111 | main | 12 | 5 | 0 |
| P112 | P112 | main | 12 | 0 | 0 |
| P113 | P113 | main | 11 | 5 | 0 |
| P114 | P114 | main | 14 | 2 | 0 |
| P115 | P115 | main | 13 | 2 | 0 |
| P116 | P116 | main | 12 | 4 | 0 |
| P117 | P117 | main | 10 | 4 | 0 |
| P118 | P118 | main | 18 | 6 | 0 |
| P119 | P119 | main | 7 | 0 | 1 |
| P120 | P120 | main | 14 | 1 | 0 |
| P121 | P121 | main | 16 | 16 | 0 |
| P122 | P122 | main | 21 | 13 | 1 |
| P123 | P123 | main | 32 | 6 | 1 |
| P124 | P124 | main | 24 | 15 | 0 |
| P125 | P125 | main | 8 | 0 | 0 |
| P126 | P126 | main | 16 | 1 | 0 |
| P127 | P127 | main | 7 | 4 | 0 |
| P128 | P128 | main | 20 | 1 | 3 |
| P129 | P129 | main | 16 | 8 | 0 |
| P130 | P130 | main | 13 | 2 | 0 |
| P131 | P131 | main | 26 | 4 | 0 |
| P132 | P132 | main | 10 | 2 | 0 |
| P133 | P133 | main | 8 | 2 | 0 |
| P134 | P134 | main | 12 | 8 | 0 |
| P135 | P135 | main | 10 | 2 | 2 |
| P136 | P136 | main | 17 | 6 | 0 |
| P137 | P137 | main | 14 | 6 | 0 |
| P138 | P138 | main | 7 | 0 | 0 |
| P139 | P139 | main | 11 | 1 | 0 |
| P140 | P140 | main | 17 | 4 | 0 |
| P141 | P141 | main | 18 | 4 | 0 |
| P142 | P142 | main | 7 | 3 | 0 |
| P143 | P143 | main | 22 | 9 | 1 |
| P144 | P144 | main | 54 | 31 | 1 |
| P145 | P145 | main | 10 | 3 | 0 |
| P146 | P146 | main | 24 | 3 | 1 |
| P147 | P147 | main | 30 | 3 | 0 |
| P148 | P148 | main | 18 | 0 | 0 |
| P149 | P149 | main | 14 | 6 | 0 |
| P150 | P150 | main | 17 | 1 | 0 |
| P151 | P151 | main | 118 | 3 | 1 |
| P152 | P152 | main | 30 | 2 | 0 |
| P153 | P153 | main | 26 | 3 | 0 |
| P154 | P154 | main | 10 | 7 | 0 |
| P155 | P155 | main | 15 | 7 | 0 |
| P156 | P156 | main | 24 | 6 | 1 |
| P157 | P157 | main | 14 | 5 | 0 |
| P158 | P158 | main | 108 | 13 | 8 |
| P159 | P159 | main | 100 | 26 | 1 |
| P160 | P160 | main | 41 | 10 | 1 |
| P161 | P161 | main | 44 | 4 | 0 |
| P162 | P162 | main | 36 | 9 | 0 |
| P163 | P163 | main | 42 | 3 | 0 |
| P164 | P164 | main | 16 | 3 | 1 |
| P165 | P165 | main | 11 | 0 | 0 |
| P166 | P166 | main | 17 | 6 | 1 |
| P167 | P167 | main | 23 | 10 | 0 |
| P168 | P168 | main | 9 | 0 | 0 |
| P169 | P169 | main | 6 | 1 | 0 |
| P170 | P170 | main | 30 | 3 | 0 |
| P171 | P171 | main | 21 | 3 | 1 |
| P172 | P172 | main | 30 | 3 | 0 |
| P173 | P173 | main | 24 | 1 | 1 |
| P174 | P174 | main | 53 | 7 | 1 |
| P175 | P175 | main | 30 | 7 | 0 |
| P176 | P176 | main | 29 | 2 | 0 |
| P177 | P177 | main | 17 | 3 | 2 |
| P178 | P178 | main | 25 | 4 | 0 |
| P179 | P179 | main | 37 | 3 | 0 |
| P180 | P180 | main | 36 | 6 | 4 |
| P181 | P181 | main | 15 | 6 | 0 |
| P182 | P182 | main | 22 | 6 | 0 |
| P183 | P183 | main | 41 | 6 | 6 |
| P184 | P184 | main | 13 | 0 | 0 |
| P185 | P185 | main | 5 | 2 | 0 |
| P186 | P186 | main | 37 | 0 | 0 |
| P187 | P187 | main | 16 | 0 | 0 |
| P188 | P188 | main | 63 | 3 | 5 |
| P189 | P189 | main | 9 | 3 | 1 |
| P190 | P190 | main | 31 | 3 | 0 |
| P191 | P191 | main | 13 | 5 | 0 |
| P191-App | P191 | appendix | 19 | 5 | 0 |
| P192 | P192 | main | 7 | 0 | 1 |
| P193 | P193 | main | 37 | 10 | 6 |
| P194 | P194 | main | 38 | 0 | 0 |
| P195 | P195 | main | 20 | 4 | 1 |
| P196 | P196 | main | 35 | 7 | 0 |
| P197 | P197 | main | 29 | 10 | 4 |
| P198 | P198 | main | 16 | 2 | 0 |
| P199 | P199 | main | 41 | 0 | 1 |
| C018 | P224 | main | 8 | 3 | 0 |
| C019 | P242 | main | 29 | 5 | 0 |
| C020 | P211 | main | 28 | 1 | 6 |
| C021 | P161 | main | 33 | 0 | 0 |
| C022 | P215 | main | 35 | 12 | 1 |
| C024 | P241 | main | 19 | 1 | 0 |
| C025 | P219 | main | 30 | 3 | 0 |
| C026 | P207 | main | 15 | 3 | 1 |
| C030 | P232 | main | 25 | 0 | 0 |
| C031 | P230 | main | 18 | 5 | 0 |
| C037 | P202 | main | 28 | 13 | 2 |
| C038 | P210 | main | 37 | 3 | 0 |
| C041 | P205 | main | 43 | 11 | 2 |
| C042 | P209 | main | 7 | 0 | 0 |
| C043 | P204 | main | 33 | 5 | 1 |
| C062 | P214 | main | 58 | 8 | 2 |
| C063 | P221 | main | 18 | 4 | 0 |
| C065 | P236 | main | 26 | 4 | 2 |
| C068 | P217 | main | 30 | 2 | 0 |
| C070 | P206 | main | 19 | 2 | 4 |
| C071 | P251 | main | 38 | 16 | 0 |
| C072 | P203 | main | 45 | 8 | 0 |
| C073 | P252 | main | 21 | 4 | 0 |
| C074 | P237 | main | 20 | 2 | 0 |
| C075 | P201 | main | 16 | 5 | 0 |
| C076 | P223 | main | 19 | 3 | 0 |
| C078 | P212 | main | 5 | 3 | 0 |
| C080 | P247 | main | 60 | 9 | 1 |
| C081 | P218 | main | 17 | 4 | 0 |
| C082 | P227 | main | 67 | 3 | 0 |
| C083 | P235 | main | 64 | 33 | 0 |
| C084 | P222 | main | 25 | 5 | 0 |
| C087 | P228 | main | 37 | 10 | 3 |
| C090 | P234 | main | 12 | 4 | 0 |
| C093 | P226 | main | 19 | 0 | 3 |
| C095 | P244 | main | 9 | 2 | 0 |
| C130 | P233 | main | 38 | 4 | 0 |
| C137 | P250 | main | 55 | 23 | 3 |
| C156 | P208 | main | 14 | 0 | 0 |
| C160 | P225 | main | 18 | 0 | 0 |
| C170 | P243 | main | 30 | 4 | 2 |
| C217 | P240 | main | 24 | 4 | 0 |
| C261 | P238 | main | 12 | 6 | 0 |
| C264 | P246 | main | 12 | 0 | 0 |
| C266 | P220 | main | 10 | 0 | 0 |
| C269 | P216 | main | 17 | 2 | 2 |
| C276 | P213 | main | 12 | 7 | 0 |
| C279 | P253 | main | 12 | 1 | 1 |
| C281 | P248 | main | 9 | 2 | 0 |
| C286 | P239 | main | 13 | 4 | 0 |
| C260 | P238 | appendix | 3 | 0 | 1 |
| C268 | P216 | appendix | 5 | 0 | 1 |
| C275 | P213 | appendix | 9 | 4 | 1 |
| C271 | P245 | main | 14 | 2 | 0 |
| C272 | P229 | main | 12 | 0 | 0 |
| C274 | P231 | main | 15 | 5 | 0 |
| C085 | P249 | main | 21 | 1 | 1 |
| C089 | P200 | main | 47 | 5 | 0 |
| C016 | P077 | appendix | 7 | 0 | 0 |
| C017 | P082 | appendix | 14 | 1 | 7 |
| C263 | P246 | appendix | 21 | 0 | 0 |
| C278 | P253 | appendix | 25 | 9 | 1 |
## 01 Theory and Hypotheses
### P058 - Fig. 1
**Paper:** Can inquiry-based pedagogy improve math performance? Evidence from 5711 students in Vietnam on the mediating role of math attitude
**Original caption:** Fig. 1. : Diagram of conceptual relationship and Hypotheses 1, 2, 3 and 4 on the interlink between inquiry-based pedagogy (IBP), student math attitude (SMA) and student math performance (SMP).
**Purpose:** Hypothesized inquiry-pedagogy, attitude and performance paths. **Visual forms:** Path diagram.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P058-Fig01-Diagram%20of%20conceptual%20relationship%20and%20Hypotheses%201,%202,%203%20and%204%20on%20the%20interlink%20between%20inquiry-based%20pedagogy%20(IBP),%20student%20math%20attitude%20(SMA)%20and%20student%20math....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1>) · Original paper and literature note are not hosted.
### P059 - Fig. 1
**Paper:** Creative learning environments and mathematics self-efficacy as predictors of mathematics achievement: Insights from PISA 2022 across three countries – Korea, Canada, and Türkiye
**Original caption:** Fig. 1. The hypothesized model.
**Purpose:** Hypothesized creative-environment and self-efficacy relationships. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P059-Fig01-The%20hypothesized%20model.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=2>) · Original paper and literature note are not hosted.
### P060 - Fig. 1
**Paper:** Does motivation towards science mediate the relationship between parental support and adolescent science achievement? A comparison of two countries
**Original caption:** Fig. 1. Conceptual Framework.
**Purpose:** Conceptual mediating and country-dependent relationships. **Visual forms:** Path diagram.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P060-Fig01-Conceptual%20Framework.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=3>) · Original paper and literature note are not hosted.
### P061 - Fig. 1
**Paper:** Does the relationship between competitiveness and student achievement vary based on the culture of competitiveness? Evidence from PISA 2018
**Original caption:** Fig. 1. Adaptation of the Multilevel model of competition and competitiveness. Note. Adapted from “Steel sharpens steel: A review of multilevel of competition and competitiveness in organizations,” by Swab, R. G., & Johnson, P. D., 2019, Journal of Organizational Behavior, 40(2), p.150.
**Purpose:** Adapted multilevel competition framework motivates hypotheses. **Visual forms:** Conceptual framework.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P061-Fig01-Adaptation%20of%20the%20Multilevel%20model%20of%20competition%20and%20competitiveness.%20Note.%20Adapted%20from%20%E2%80%9CSteel%20sharpens%20steel%20A%20review%20of%20multilevel%20of%20competition%20and%20competitiveness....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=4>) · Original paper and literature note are not hosted.
### P061 - Fig. 2
**Paper:** Does the relationship between competitiveness and student achievement vary based on the culture of competitiveness? Evidence from PISA 2018
**Original caption:** Fig. 2. Conceptual model.
**Purpose:** Hypothesized cross-level competition relations. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P061-Fig02-Conceptual%20model.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=5>) · Original paper and literature note are not hosted.
### P062 - Fig. 1
**Paper:** PISA 2022 insights on intellectual curiosity, perspective-taking, and science achievement: Examining the mediating pathways
**Original caption:** Fig. 1. The structural model.
**Purpose:** Hypothesized mediation model linking curiosity, perspective-taking, and science. **Visual forms:** Path diagram.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P062-Fig01-The%20structural%20model.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=6>) · Original paper and literature note are not hosted.
### P064 - Fig. 1
**Paper:** Students’ reading enjoyment as a multilevel mediator of the relationship between parental reading enjoyment, teachers’ stimulation of reading, ICT factors, and reading literacy
**Original caption:** Fig. 1. The research model.
**Purpose:** Hypothesized student- and school-level mediation. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 5.
[PNG](<01_Theory_and_Hypotheses/P064-Fig01-The%20research%20model.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=7>) · Original paper and literature note are not hosted.
### P066 - Fig. 1
**Paper:** The impact of financial education in schools on the financial literacy of Brazilian students
**Original caption:** Fig. 1. – Rates of return on investment in human capital. Source: Adapted from Heckman (2006).
**Purpose:** Human-capital return curve supplies the theoretical rationale for financial education. **Visual forms:** Conceptual curve.
**Physical source pages:** 2.
[PNG](<01_Theory_and_Hypotheses/P066-Fig01-Rates%20of%20return%20on%20investment%20in%20human%20capital.%20Source%20Adapted%20from%20Heckman%20(2006).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=8>) · Original paper and literature note are not hosted.
### P069 - Fig. 1
**Paper:** What students learn depends on what we teach: Curriculum content coverage and student achievement in mathematics and science
**Original caption:** Fig. 1. OTL Curriculum Model. Source: Inspired by Exhibit 1 in Mullis and Martin (2017).
**Purpose:** Opportunity-to-learn conceptual curriculum framework. **Visual forms:** Flowchart.
**Physical source pages:** 2.
[PNG](<01_Theory_and_Hypotheses/P069-Fig01-OTL%20Curriculum%20Model.%20Source%20Inspired%20by%20Exhibit%201%20in%20Mullis%20and%20Martin%20(2017).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=9>) · Original paper and literature note are not hosted.
### P074 - Fig. 1
**Paper:** Predicting Key Factors Influencing Students’ Creative Thinking: A Machine Learning Analysis Based on PISA 2022 Data from East Asia
**Original caption:** Fig. 1. Competency model for the PISA test of creative thinking.
**Purpose:** Defines the PISA creative-thinking competency construct. **Visual forms:** Composite.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P074-Fig01-Competency%20model%20for%20the%20PISA%20test%20of%20creative%20thinking.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=10>) · Original paper and literature note are not hosted.
### P074 - Fig. 2
**Paper:** Predicting Key Factors Influencing Students’ Creative Thinking: A Machine Learning Analysis Based on PISA 2022 Data from East Asia
**Original caption:** Fig. 2. Enablers and manifestations of creative thinking in the classroom.
**Purpose:** Conceptual framework of creative-thinking enablers and manifestations. **Visual forms:** Composite.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P074-Fig02-Enablers%20and%20manifestations%20of%20creative%20thinking%20in%20the%20classroom.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=11>) · Original paper and literature note are not hosted.
### P074 - Fig. 3
**Paper:** Predicting Key Factors Influencing Students’ Creative Thinking: A Machine Learning Analysis Based on PISA 2022 Data from East Asia
**Original caption:** Fig. 3. Socio-ecological theory.
**Purpose:** Socio-ecological theory informing predictor selection. **Visual forms:** Composite.
**Physical source pages:** 5.
[PNG](<01_Theory_and_Hypotheses/P074-Fig03-Socio-ecological%20theory.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=12>) · Original paper and literature note are not hosted.
### P074 - Fig. 4
**Paper:** Predicting Key Factors Influencing Students’ Creative Thinking: A Machine Learning Analysis Based on PISA 2022 Data from East Asia
**Original caption:** Fig. 4. Model of factors influencing creative thinking.
**Purpose:** Framework organizes hypothesized predictors by ecological systems. **Visual forms:** Path diagram.
**Physical source pages:** 5.
[PNG](<01_Theory_and_Hypotheses/P074-Fig04-Model%20of%20factors%20influencing%20creative%20thinking.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=13>) · Original paper and literature note are not hosted.
### P078 - Fig. 1
**Paper:** Digital resource engagement, health challenges, and mathematical achievement in school children: An in-depth analysis of PISA data 2022
**Original caption:** Fig. 1. Conceptual framework (hypothesized model).
**Purpose:** Hypothesized structural relations among resource use, health, confidence, and achievement. **Visual forms:** Path diagram.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P078-Fig01-Conceptual%20framework%20(hypothesized%20model).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=14>) · Original paper and literature note are not hosted.
### P079 - Fig. 1
**Paper:** Evaluating the effects of ICT engagement on science and mathematics achievement: Evidence from PISA 2022 using structural equation modeling
**Original caption:** Fig. 1. Conceptual framework. Notes: HM: ICT available outside school (ICTHOME), OUT: Use of ICT for school activities of the classroom (ICTOUT), SCH: ICT available at school (ICTSCH), LEI: Leisure activities using ICT (ICTWKDY), SE: Self-efficacy in digital competences (ICTEFFIC).
**Purpose:** Conceptual model linking ICT engagement dimensions, digital self-efficacy, and achievement. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P079-Fig01-Conceptual%20framework.%20Notes%20HM%20ICT%20available%20outside%20school%20(ICTHOME),%20OUT%20Use%20of%20ICT%20for%20school%20activities%20of%20the%20classroom%20(ICTOUT),%20SCH%20ICT%20available%20at%20school....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=15>) · Original paper and literature note are not hosted.
### P080 - Fig. 1
**Paper:** Examining the relationship between online chatting and PISA reading literacy trends (2000-2018)
**Original caption:** Fig. 1. Conceptual model.
**Purpose:** Conceptual model motivating country-level trend relationships. **Visual forms:** Path diagram.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P080-Fig01-Conceptual%20model.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=16>) · Original paper and literature note are not hosted.
### P083 - Fig. 1
**Paper:** Profiles of academic performance and perceived digital distraction in secondary classrooms
**Original caption:** Fig. 1. The theoretical framework of the present study based on social cognitive theory.
**Purpose:** Social cognitive framework organizes predictors of distraction-performance profiles. **Visual forms:** Conceptual framework.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P083-Fig01-The%20theoretical%20framework%20of%20the%20present%20study%20based%20on%20social%20cognitive%20theory.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=17>) · Original paper and literature note are not hosted.
### P085 - Fig. 1
**Paper:** The impact of digital resource utilization on student learning outcomes and self-efficacy across different economic contexts: A comparative analysis of PISA, 2022
**Original caption:** Fig. 1. Conceptual framework.
**Purpose:** Hypothesized digital-use and efficacy paths. **Visual forms:** Path diagram.
**Physical source pages:** 5.
[PNG](<01_Theory_and_Hypotheses/P085-Fig01-Conceptual%20framework.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=18>) · Original paper and literature note are not hosted.
### P088 - Fig. 1
**Paper:** Understanding the relationships among ICT use, self-efficacy, and achievement in PISA 2022: A multigroup analysis featuring gender and immigrant status
**Original caption:** Fig. 1. The hypothesized model of the relationships among ICT use, self-efficacy in digital competencies, and achievement.
**Purpose:** Hypothesized ICT–self-efficacy–achievement mediation framework. **Visual forms:** Path diagram.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P088-Fig01-The%20hypothesized%20model%20of%20the%20relationships%20among%20ICT%20use,%20self-efficacy%20in%20digital%20competencies,%20and%20achievement.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=19>) · Original paper and literature note are not hosted.
### P090 - Fig. 1
**Paper:** Meta-affective progression in science: A cross-sectional study of junior high school students
**Original caption:** Fig. 1. Indicators and sub-indicators of meta-affective according to Kondacki (2016).
**Purpose:** Conceptual hierarchy defines meta-affective awareness and regulation. **Visual forms:** Flowchart.
**Physical source pages:** 2.
[PNG](<01_Theory_and_Hypotheses/P090-Fig01-Indicators%20and%20sub-indicators%20of%20meta-affective%20according%20to%20Kondacki%20(2016).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=20>) · Original paper and literature note are not hosted.
### P093 - Table 1
**Paper:** Using machine learning algorithms to predict students’ general self-efficacy in PISA 2018
**Original caption:** Table 1 The Classification of Contextual Predictors of Self-efficacy. Table 1 (continued)
**Purpose:** Multilevel ecological taxonomy links conceptual domains and predictor modules. **Visual forms:** Complex table.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P093-Table01-The%20Classification%20of%20Contextual%20Predictors%20of%20Self-efficacy.%20Table%201%20(continued).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=21>) · Original paper and literature note are not hosted.
### P095 - Fig. 1
**Paper:** International differences in educational equity: An assessment using the Benefit of the Doubt model
**Original caption:** Fig. 1. Dimensions and indicators for the equity composite indicator.
**Purpose:** Conceptual dimensions underlying the equity composite. **Visual forms:** Flowchart.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P095-Fig01-Dimensions%20and%20indicators%20for%20the%20equity%20composite%20indicator.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=22>) · Original paper and literature note are not hosted.
### P096 - Fig. 2
**Paper:** Pathways to equity: A mediation analysis of gender, SES, and mathematics achievement using PISA 2022 UK data
**Original caption:** Fig. 2. Hypothesis diagram of mediational analyses among student individual characteristics, student learning process, and mathematics achievement.
**Purpose:** Hypothesized mediation model. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P096-Fig02-Hypothesis%20diagram%20of%20mediational%20analyses%20among%20student%20individual%20characteristics,%20student%20learning%20process,%20and%20mathematics%20achievement.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=23>) · Original paper and literature note are not hosted.
### P097 - Fig. 1
**Paper:** The role of parental involvement in adolescents’ digital reading performance
**Original caption:** Fig. 1. Parental involvement.
**Purpose:** Conceptual operational framework for parental behavioral/emotional involvement and their interaction. **Visual forms:** Path diagram.
**Physical source pages:** 2.
[PNG](<01_Theory_and_Hypotheses/P097-Fig01-Parental%20involvement.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=24>) · Original paper and literature note are not hosted.
### P102 - Fig. 1
**Paper:** Fostering school and classroom discipline in Türkiye: Leadership, shared decision-making, parental involvement and teacher self-efficacy
**Original caption:** Fig. 1. Theoretical and conceptual model depicting the relationships among study variables. Source: Authors own works.
**Purpose:** Ecological theoretical model of leadership and classroom discipline. **Visual forms:** Path diagram.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P102-Fig01-Theoretical%20and%20conceptual%20model%20depicting%20the%20relationships%20among%20study%20variables.%20Source%20Authors%20own%20works.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=25>) · Original paper and literature note are not hosted.
### P103 - Fig. 1
**Paper:** Impact of school leadership on teachers’ instructional practices: examining mediating roles of collegiality and teacher autonomy using Vietnam TALIS 2018
**Original caption:** Fig. 1. Conceptual framework.
**Purpose:** Conceptual leadership–collegiality–autonomy–instructional-practice model. **Visual forms:** Path diagram.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P103-Fig01-Conceptual%20framework.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=26>) · Original paper and literature note are not hosted.
### P105 - Fig. 1
**Paper:** Overworked and under-creative: Teacher workload, time allocation, and creative pedagogy in Chinese secondary schools - evidence from PISA 2022
**Original caption:** Fig. 1. Research framework.
**Purpose:** Multilevel conceptual framework links time management, resources and creative pedagogy. **Visual forms:** Path diagram.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P105-Fig01-Research%20framework.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=27>) · Original paper and literature note are not hosted.
### P108 - Fig. 1
**Paper:** Understanding the relationship between teacher collaboration and instructional clarity via teacher self-efficacy: A moderated mediation model of organizational commitment
**Original caption:** Fig. 1. Analytical Framework and Hypotheses.
**Purpose:** Hypotheses for teacher collaboration, efficacy, clarity and organizational commitment. **Visual forms:** Path diagram.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P108-Fig01-Analytical%20Framework%20and%20Hypotheses.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=28>) · Original paper and literature note are not hosted.
### P111 - Fig. 1
**Paper:** Distributed leadership and teacher job satisfaction over time: A large-scale trend analysis of TALIS 2013 and 2018
**Original caption:** Fig. 1. Path models of the relationship between distributed leadership and teacher job satisfaction in TALIS 2013 and 2018.
**Purpose:** Specifies the two-wave leadership and job-satisfaction path models. **Visual forms:** Path diagram.
**Physical source pages:** 5.
[PNG](<01_Theory_and_Hypotheses/P111-Fig01-Path%20models%20of%20the%20relationship%20between%20distributed%20leadership%20and%20teacher%20job%20satisfaction%20in%20TALIS%202013%20and%202018.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=29>) · Original paper and literature note are not hosted.
### P113 - Fig. 1
**Paper:** How perceived teaching autonomy predicts job satisfaction in China: The mediating role of self-efficacy and the moderating role of school climate
**Original caption:** Fig. 1. Moderated mediation model.
**Purpose:** Hypothesized moderated-mediation model. **Visual forms:** Path diagram.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P113-Fig01-Moderated%20mediation%20model.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=30>) · Original paper and literature note are not hosted.
### P114 - Fig. 1
**Paper:** Investigating the relationship between instructional leadership and teacher well-being: A moderated mediation analysis of PISA 2022
**Original caption:** Fig. 1. A conceptual framework of the relationships among perceived instructional leadership, teacher workload stress, teacher autonomy support, and teachers’ well-being.
**Purpose:** JD-R conceptual moderated-mediation framework. **Visual forms:** Path diagram.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P114-Fig01-A%20conceptual%20framework%20of%20the%20relationships%20among%20perceived%20instructional%20leadership,%20teacher%20workload%20stress,%20teacher%20autonomy%20support,%20and%20teachers%E2%80%99%20well-being.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=31>) · Original paper and literature note are not hosted.
### P117 - Fig. 1
**Paper:** The effect of teacher collaboration on teachers' career well-being in China: A moderated mediation model of teacher self-efficacy and distributed leadership
**Original caption:** Fig. 1. Research model.
**Purpose:** Hypothesized moderated mediation model. **Visual forms:** Path diagram.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P117-Fig01-Research%20model.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=32>) · Original paper and literature note are not hosted.
### P118 - Fig. 1
**Paper:** Understanding the relationship between teacher collaboration beliefs and job satisfaction via collaborative attitudes: A moderated mediation model of classroom self-efficacy from PISA 2022
**Original caption:** Fig. 1. Analytical framework and hypotheses.
**Purpose:** Hypothesized collaboration-belief–attitude–satisfaction mediation and efficacy moderation. **Visual forms:** Path diagram.
**Physical source pages:** 5.
[PNG](<01_Theory_and_Hypotheses/P118-Fig01-Analytical%20framework%20and%20hypotheses.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=33>) · Original paper and literature note are not hosted.
### P120 - Fig. 1
**Paper:** National policy, school contexts, and individual characteristics: Which matters most for teachers’ participation in induction programs?
**Original caption:** Fig. 1. Conceptual framework. Note: PD record is the school’s professional development records and documentation.
**Purpose:** Country, school and teacher determinants in the hypothesized induction-participation framework. **Visual forms:** Path diagram.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P120-Fig01-Conceptual%20framework.%20Note%20PD%20record%20is%20the%20school%E2%80%99s%20professional%20development%20records%20and%20documentation.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=34>) · Original paper and literature note are not hosted.
### P122 - Fig. 1
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 1. Teacher assessment literacy conceptual model (Pastore & Andrade, 2019).
**Purpose:** Research framework. **Visual forms:** Conceptual model.
**Physical source pages:** 2.
[PNG](<01_Theory_and_Hypotheses/P122-Fig01-Teacher%20assessment%20literacy%20conceptual%20model%20(Pastore%20&%20Andrade,%202019).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=35>) · Original paper and literature note are not hosted.
### P123 - Fig. 1
**Paper:** Professional Development of New Higher Education Teachers With Information and Communication Technology in Shanghai: A Kirkpatrick’s Evaluation Approach
**Original caption:** Figure 1. Kirkpatrick’s evaluation model (Kirkpatrick, 1998).
**Purpose:** Frameworks used to define and evaluate technology-supported teacher development. **Visual forms:** Conceptual model, Path diagram.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P123-Fig01-Kirkpatrick%E2%80%99s%20evaluation%20model%20(Kirkpatrick,%201998).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=36>) · Original paper and literature note are not hosted.
### P123 - Fig. 2
**Paper:** Professional Development of New Higher Education Teachers With Information and Communication Technology in Shanghai: A Kirkpatrick’s Evaluation Approach
**Original caption:** Figure 2. Extended technology acceptance model adapted from Davis (1989).
**Purpose:** Frameworks used to define and evaluate technology-supported teacher development. **Visual forms:** Conceptual model, Path diagram.
**Physical source pages:** 5.
[PNG](<01_Theory_and_Hypotheses/P123-Fig02-Extended%20technology%20acceptance%20model%20adapted%20from%20Davis%20(1989).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=37>) · Original paper and literature note are not hosted.
### P123 - Fig. 3
**Paper:** Professional Development of New Higher Education Teachers With Information and Communication Technology in Shanghai: A Kirkpatrick’s Evaluation Approach
**Original caption:** Figure 3. TPACK framework (Mishra & Koehler, 2011).
**Purpose:** Frameworks used to define and evaluate technology-supported teacher development. **Visual forms:** Conceptual model, Path diagram.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P123-Fig03-TPACK%20framework%20(Mishra%20&%20Koehler,%202011).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=38>) · Original paper and literature note are not hosted.
### P123 - Fig. 4
**Paper:** Professional Development of New Higher Education Teachers With Information and Communication Technology in Shanghai: A Kirkpatrick’s Evaluation Approach
**Original caption:** Figure 4. Teacher professional development impacts. (Steinert et al., 2006).
**Purpose:** Frameworks used to define and evaluate technology-supported teacher development. **Visual forms:** Conceptual model, Path diagram.
**Physical source pages:** 7.
[PNG](<01_Theory_and_Hypotheses/P123-Fig04-Teacher%20professional%20development%20impacts.%20(Steinert%20et%20al.,%202006).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=39>) · Original paper and literature note are not hosted.
### P127 - Fig. 1
**Paper:** Teacher evaluation – A resource for teaching quality or a burden? An analysis using TALIS 2018 data
**Original caption:** Fig. 1. Simplified model of the effects of teacher evaluation.
**Purpose:** Simplified theoretical model of evaluation and outcomes. **Visual forms:** Path diagram.
**Physical source pages:** 2.
[PNG](<01_Theory_and_Hypotheses/P127-Fig01-Simplified%20model%20of%20the%20effects%20of%20teacher%20evaluation.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=40>) · Original paper and literature note are not hosted.
### P127 - Fig. 2
**Paper:** Teacher evaluation – A resource for teaching quality or a burden? An analysis using TALIS 2018 data
**Original caption:** Fig. 2. Theoretical model of the effects of teacher evaluation.
**Purpose:** Detailed hypothesized teaching-quality dimensions. **Visual forms:** Path diagram.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P127-Fig02-Theoretical%20model%20of%20the%20effects%20of%20teacher%20evaluation.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=41>) · Original paper and literature note are not hosted.
### P133 - Fig. 1
**Paper:** How justified are the criticisms of bias against the OECD’s global competence framework?
**Original caption:** Fig. 1. Global competence dimensions in PISA 2018 (OECD, 2018, p.22).
**Purpose:** Defines the global-competence framework under evaluation. **Visual forms:** Conceptual framework.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P133-Fig01-Global%20competence%20dimensions%20in%20PISA%202018%20(OECD,%202018,%20p.22).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=42>) · Original paper and literature note are not hosted.
### P134 - Fig. 1
**Paper:** Separating prior knowledge from acquired knowledge: An individual differences analysis of PISA - learning in the digital world
**Original caption:** Fig. 1. Expected structural relations to validate the interpretation of the residual variable as “acquired knowledge and skills”.
**Purpose:** Expected relations frame the residual-validity hypotheses. **Visual forms:** Path diagram.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P134-Fig01-Expected%20structural%20relations%20to%20validate%20the%20interpretation%20of%20the%20residual%20variable%20as%20%E2%80%9Cacquired%20knowledge%20and%20skills%E2%80%9D.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=43>) · Original paper and literature note are not hosted.
### P141 - Fig. 1
**Paper:** Unraveling the nexus: Culture, cognitive competence, and economic performance across 86 nations (2000–2018)
**Original caption:** Fig. 1. Axiological cube of three cultures: honor, achievement, and joy. Note. Model taken from Basáñez (2016, p. 71); Axiological dimensions: A to B = grew richer from abundant agriculture, A to C = increased power as population and pyramidal hierarchies grew, A to E = social differentiation and specialization advanced; Axiological walls: 1.1 = Hierarchy, 1.2 = Egalitarianism, 2.1 = Harmony, 2.2 = Mastery, 3.1 = Embeddedness, 3.2 = Autonomy.
**Purpose:** Axiological theory of honor, achievement and joy cultures. **Visual forms:** Composite.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P141-Fig01-Axiological%20cube%20of%20three%20cultures%20honor,%20achievement,%20and%20joy.%20Note.%20Model%20taken%20from%20Bas%C3%A1%C3%B1ez%20(2016,%20p.%2071);%20Axiological%20dimensions%20A%20to%20B%20=%20grew%20richer%20from%20abundant....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=44>) · Original paper and literature note are not hosted.
### P141 - Fig. 2
**Paper:** Unraveling the nexus: Culture, cognitive competence, and economic performance across 86 nations (2000–2018)
**Original caption:** Fig. 2. Basic theoretical model for interactions between national culture, student performance and economy. Note. Dotted line = possible feedback.
**Purpose:** Theoretical relationship between culture, student competence and economy. **Visual forms:** Path diagram.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P141-Fig02-Basic%20theoretical%20model%20for%20interactions%20between%20national%20culture,%20student%20performance%20and%20economy.%20Note.%20Dotted%20line%20=%20possible%20feedback.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=45>) · Original paper and literature note are not hosted.
### P141 - Fig. 3
**Paper:** Unraveling the nexus: Culture, cognitive competence, and economic performance across 86 nations (2000–2018)
**Original caption:** Fig. 3. Theoretical path model for direct and mediated effects of national culture on economic growth. Note. Solid arrows represent the path model without controls (left half of Table 7 [I]), dashed arrows are additionally added as controls (right half of Table 7 [II]); results for effects Dir., A, B and mediated AB are given in Table 7.
**Purpose:** Hypothesized direct and mediated cultural associations with economic growth. **Visual forms:** Path diagram.
**Physical source pages:** 7.
[PNG](<01_Theory_and_Hypotheses/P141-Fig03-Theoretical%20path%20model%20for%20direct%20and%20mediated%20effects%20of%20national%20culture%20on%20economic%20growth.%20Note.%20Solid%20arrows%20represent%20the%20path%20model%20without%20controls%20(left%20half%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=46>) · Original paper and literature note are not hosted.
### P147 - Fig. 1
**Paper:** A Review of Generalizability and Transportability
**Original caption:** Figure 1. Internal vs. external validity biases as they relate to target, study, and analysis populations.
**Purpose:** Conceptual map distinguishes internal/external validity across populations and samples. **Visual forms:** Flowchart.
**Physical source pages:** 2.
[PNG](<01_Theory_and_Hypotheses/P147-Fig01-Internal%20vs.%20external%20validity%20biases%20as%20they%20relate%20to%20target,%20study,%20and%20analysis%20populations.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=47>) · Original paper and literature note are not hosted.
### P147 - Fig. 3
**Paper:** A Review of Generalizability and Transportability
**Original caption:** Figure 3. Illustrative example of the difference between target population and sample average treatment effects (PATE and SATE). Biases may differ in magnitude and may make the SATE either larger or smaller than the PATE.
**Purpose:** Illustrative estimand decomposition explains the relation between target and sample average effects. **Visual forms:** Conceptual diagram.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P147-Fig03-Illustrative%20example%20of%20the%20difference%20between%20target%20population%20and%20sample%20average%20treatment%20effects%20(PATE%20and%20SATE).%20Biases%20may%20differ%20in%20magnitude%20and%20may%20make%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=48>) · Original paper and literature note are not hosted.
### P150 - Fig. 1
**Paper:** Data quality disparities in large-scale assessments: insufficient effort responding across student groups, schools, and cultures
**Original caption:** Fig. 1 Three-level ecosystem of responding motivation
**Purpose:** Theoretical ecological layers of survey-response motivation. **Visual forms:** Conceptual framework.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P150-Fig01-Three-level%20ecosystem%20of%20responding%20motivation.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=49>) · Original paper and literature note are not hosted.
### P155 - Fig. 1
**Paper:** Educational participation, cultural resources, and social inequality: Have social inequalities in reading literacy increased or decreased between 2000 and 2022?
**Original caption:** Figure 1. Analytical framework and theoretical expectations. Notes. Hypotheses H1 and H2 are divided into the partial aspects of whether learning environments differ according to social background (distribution effect: H1a; H2a) and to what extent these learning environments are related to the acquisition of reading skills (correlation effect: H1b; H2b).
**Purpose:** Hypothesized distributional and associational mechanisms of educational inequality. **Visual forms:** Path diagram.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P155-Fig01-Analytical%20framework%20and%20theoretical%20expectations.%20Notes.%20Hypotheses%20H1%20and%20H2%20are%20divided%20into%20the%20partial%20aspects%20of%20whether%20learning%20environments%20differ%20according%20to....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=50>) · Original paper and literature note are not hosted.
### P158 - Fig. 2
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 2 ISAR model based on Bauer et al. (2025)
**Purpose:** ISAR theoretical model relates AI use to cognitive engagement. **Visual forms:** Conceptual framework.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P158-Fig02-ISAR%20model%20based%20on%20Bauer%20et%20al.%20(2025).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=51>) · Original paper and literature note are not hosted.
### P158 - Fig. 10
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 10 Illustration of the influence of inversion arising from challenges of learners on each ISAR level
**Purpose:** Conceptual account of learner challenges reversing expected engagement benefits. **Visual forms:** Conceptual framework.
**Physical source pages:** 35.
[PNG](<01_Theory_and_Hypotheses/P158-Fig10-Illustration%20of%20the%20influence%20of%20inversion%20arising%20from%20challenges%20of%20learners%20on%20each%20ISAR%20level.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=52>) · Original paper and literature note are not hosted.
### P158 - Fig. 11
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 11 Hypothesized conceptual framework based on the findings in this work. The solid line represents the effect found in this work. The dashed lines represent hypothesized effects of the learner and instructional support level on learning outcomes, which is likely to contribute to a large amount of unexplained heterogeneity
**Purpose:** Explicitly hypothesized mechanisms distinguish observed and proposed effects. **Visual forms:** Path diagram.
**Physical source pages:** 39.
[PNG](<01_Theory_and_Hypotheses/P158-Fig11-Hypothesized%20conceptual%20framework%20based%20on%20the%20findings%20in%20this%20work.%20The%20solid%20line%20represents%20the%20effect%20found%20in%20this%20work.%20The%20dashed%20lines%20represent%20hypothesized....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=53>) · Original paper and literature note are not hosted.
### P161 - Fig. 1
**Paper:** Extramural ICT factors impact adolescents’ academic performance and well-being differently: Types of self-regulated learners also matter
**Original caption:** Fig. 1 Conceptual framework of the study
**Purpose:** Conceptual ICT pathways and learner-profile moderation. **Visual forms:** Path diagram.
**Physical source pages:** 11.
[PNG](<01_Theory_and_Hypotheses/P161-Fig01-Conceptual%20framework%20of%20the%20study.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=54>) · Original paper and literature note are not hosted.
### P167 - Fig. 1
**Paper:** Imputation Strategies Under Clinical Presence: Impact on Algorithmic Fairness
**Original caption:** Figure 1: Examples of group-specific clinical presence mechanisms.
**Purpose:** Defines group-specific missingness mechanisms and their causal assumptions. **Visual forms:** Line/profile plot.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P167-Fig01-Examples%20of%20group-specific%20clinical%20presence%20mechanisms.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=55>) · Original paper and literature note are not hosted.
### P167 - Fig. 2
**Paper:** Imputation Strategies Under Clinical Presence: Impact on Algorithmic Fairness
**Original caption:** Figure 2: Directed Acyclic Graphs (DAGs) associated with the identified clinical missingness scenarios. Full circled covariates are observed, dotted ones unobserved. Y is
**Purpose:** Defines group-specific missingness mechanisms and their causal assumptions. **Visual forms:** Composite.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P167-Fig02-Directed%20Acyclic%20Graphs%20(DAGs)%20associated%20with%20the%20identified%20clinical%20missingness%20scenarios.%20Full%20circled%20covariates%20are%20observed,%20dotted%20ones%20unobserved.%20Y%20is.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=56>) · Original paper and literature note are not hosted.
### P181 - Fig. 1
**Paper:** Participation bias in the UK Biobank distorts genetic associations and downstream analyses
**Original caption:** Fig. 1 | The impact of participation bias in genetic studies. a–c, The relationships between a genetic variant (G), an exposure (X) or outcome (Y), and study participation (Z). Panel a illustrates the effect of participation bias in GWA studies, where Z is a common consequence of G and Y (red dotted line). Conditioning on a common consequence (Z) induces a non-causal association between G and Y. Panels b,c illustrate the effect of participation bias in MR studies, where bias occurs if Z is a consequence of either X (b) or Y (c). Conditioning on Z induces an association between the genetic variant and confounders, thereby violating the MR assumption of exchangeability. This figure is a simplified illustration of how participation bias can impact results obtained from two commonly employed methods in genomic studies. For further examples illustrating the impact of selection bias, see Hernán et al.7 .
**Purpose:** Conceptual participation-bias mechanism. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 2.
[PNG](<01_Theory_and_Hypotheses/P181-Fig01-The%20impact%20of%20participation%20bias%20in%20genetic%20studies.%20a-c,%20The%20relationships%20between%20a%20genetic%20variant%20(G),%20an%20exposure%20(X)%20or%20outcome%20(Y),%20and%20study%20participation%20(Z)....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=57>) · Original paper and literature note are not hosted.
### P182 - Fig. 1
**Paper:** Patterns of item nonresponse behaviour to survey questionnaires are systematic and associated with genetic loci
**Original caption:** Fig. 1 | Directed acyclic graph (DAG) of potential influences linking nonresponse to the association between genetics (G) and target phenotypes (Y1 and Y2). Potential elements of nonresponse include overall item non-response behaviour (I), item-specific nonresponse (I1 and I2) and survey or study-level nonparticipation or ascertainment (S). Boxes indicate sets of traits, with paths to/from a box indicating potential paths to one or more traits in the box. GWAS aims to discover direct (α) or indirect (δ×η) associations with genetics G on phenotypes Y conditional on covariates. Analyses of observed data implicitly condition on nonresponse (I,S) and thus may be biased if that conditioning affects the expected joint distribution of genetic data, covariates and phenotypes (for example, β ≠ 0, θ ≠ 0 or γ×δ or γ×η ≠ 0). Modelling the missingness mechanism, including use of mediators as auxiliary variables, can reduce bias from nonresponse that does not depend directly on the missing value (that is, paths other than θ). The current study demonstrates association of genetics with nonresponse (that is, β and δ×γ) and considers the prospect of modelling missingness in GWAS.
**Purpose:** Conceptual DAG explains the hypothesized selection and nonresponse bias pathways. **Visual forms:** Path diagram.
**Physical source pages:** 5.
[PNG](<01_Theory_and_Hypotheses/P182-Fig01-Directed%20acyclic%20graph%20(DAG)%20of%20potential%20influences%20linking%20nonresponse%20to%20the%20association%20between%20genetics%20(G)%20and%20target%20phenotypes%20(Y1%20and%20Y2).%20Potential%20elements%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=58>) · Original paper and literature note are not hosted.
### P188 - Fig. 1
**Paper:** Socioeconomic status and student learning: insights from an umbrella review
**Original caption:** Fig. 1 Framework for Umbrella review
**Purpose:** Conceptual umbrella-review framework connects SES definitions, mediators, moderators and learning outcomes. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P188-Fig01-Framework%20for%20Umbrella%20review.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=59>) · Original paper and literature note are not hosted.
### P196 - Fig. 1
**Paper:** Universal Basic Incomes vs. Targeted Transfers: Anti-Poverty Programs in Developing Countries
**Original caption:** Figure 1: Example of Progressive Post-Tax Income Schedules With and Without a Universal Basic Income
**Purpose:** Illustrates theoretical progressive income schedules. **Visual forms:** Conceptual curve.
**Physical source pages:** 30.
[PNG](<01_Theory_and_Hypotheses/P196-Fig01-Example%20of%20Progressive%20Post-Tax%20Income%20Schedules%20With%20and%20Without%20a%20Universal%20Basic%20Income.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=60>) · Original paper and literature note are not hosted.
### P196 - Fig. 2
**Paper:** Universal Basic Incomes vs. Targeted Transfers: Anti-Poverty Programs in Developing Countries
**Original caption:** Figure 2: Example of Post-Tax Income Schedules with and Without a Universal Basic Income, With a Tax-Exempt Region
**Purpose:** Illustrates theoretical post-tax income schedules. **Visual forms:** Conceptual curve.
**Physical source pages:** 30.
[PNG](<01_Theory_and_Hypotheses/P196-Fig02-Example%20of%20Post-Tax%20Income%20Schedules%20with%20and%20Without%20a%20Universal%20Basic%20Income,%20With%20a%20Tax-Exempt%20Region.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=61>) · Original paper and literature note are not hosted.
### P200 - Fig. 1
**Paper:** An opportunity cost model of subjective effort and task performance
**Original caption:** Figure 1. A schematic diagram of the proposed opportunity cost model. The expected costs and benefits of target and nontarget tasks are estimated (top). These computations give rise to phenomenology (e.g., qualia such as frustration, boredom, flow), which, in turn, motivates the allocation of computational processes to tasks that are expected to optimize costs and benefits. This allocation determines performance, on both the target and nontarget tasks. The experienced costs and benefits then recursively feed into another iteration of the same sequence, with continued adjustment of allocation decisions, but without depletion of any physical resource.
**Purpose:** Theoretical opportunity-cost allocation model and its worked hypothetical utilities. **Visual forms:** Flowchart.
**Physical source pages:** 8.
[PNG](<01_Theory_and_Hypotheses/P200-Fig01-A%20schematic%20diagram%20of%20the%20proposed%20opportunity%20cost%20model.%20The%20expected%20costs%20and%20benefits%20of%20target%20and%20nontarget%20tasks%20are%20estimated%20(top).%20These%20computations%20give....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=718>) · Original paper and literature note are not hosted.
### P200 - Fig. 2
**Paper:** An opportunity cost model of subjective effort and task performance
**Original caption:** Figure 2. Hypothetical utilities of different actions a research participant might engage in, illustrating how opportunity costs depend on the set of actions available.
**Purpose:** Theoretical opportunity-cost allocation model and its worked hypothetical utilities. **Visual forms:** Conceptual table.
**Physical source pages:** 10.
[PNG](<01_Theory_and_Hypotheses/P200-Fig02-Hypothetical%20utilities%20of%20different%20actions%20a%20research%20participant%20might%20engage%20in,%20illustrating%20how%20opportunity%20costs%20depend%20on%20the%20set%20of%20actions%20available.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=719>) · Original paper and literature note are not hosted.
### P200 - Fig. 3
**Paper:** An opportunity cost model of subjective effort and task performance
**Original caption:** Figure 3. How hypothetical utilities of different actions might change for a research participant with the experimenter present/absent, illustrating opportunity costs and the optimal action changing in different contexts.
**Purpose:** Theoretical opportunity-cost allocation model and its worked hypothetical utilities. **Visual forms:** Conceptual table.
**Physical source pages:** 11.
[PNG](<01_Theory_and_Hypotheses/P200-Fig03-How%20hypothetical%20utilities%20of%20different%20actions%20might%20change%20for%20a%20research%20participant%20with%20the%20experimenter%20present%20absent,%20illustrating%20opportunity%20costs%20and%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=720>) · Original paper and literature note are not hosted.
### P200 - Fig. 4
**Paper:** An opportunity cost model of subjective effort and task performance
**Original caption:** Figure 4. Hypothetical utilities of dedicating computational processes to one task or dividing them between two tasks, illustrating how opportunity costs apply not just to the selection of tasks but also the allocation of processes among tasks.
**Purpose:** Recovered by direct visual comparison with source page. **Visual forms:** Conceptual table.
**Physical source pages:** 12.
[PNG](<01_Theory_and_Hypotheses/P200-Fig04-Hypothetical%20utilities%20of%20dedicating%20computational%20processes%20to%20one%20task%20or%20dividing%20them%20between%20two%20tasks,%20illustrating%20how%20opportunity%20costs%20apply%20not%20just%20to%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=721>) · Original paper and literature note are not hosted.
### P200 - Fig. 5
**Paper:** An opportunity cost model of subjective effort and task performance
**Original caption:** Figure 5. For the simple model outlined in the text, whether one should focus attention on only the highest valued action or divide attention between the two best actions, as a function of the relative utility (RU) of the next-best action and the fraction of the value (β) one gains from a task when dividing processing capacity. These two factors determine the opportunity cost, and it is better to divide attention when the opportunity cost is high. The locations x and y in (B) provide an example of how to think about the dynamics of effort and performance. A person will feel an increased sense of effort, and be motivated to reallocate attention/mental processes in a way that reduces performance on a task, when the perceived costs and benefits of the task move from position x to position y.
**Purpose:** Theoretical opportunity-cost allocation model and its worked hypothetical utilities. **Visual forms:** Conceptual curve.
**Physical source pages:** 14.
[PNG](<01_Theory_and_Hypotheses/P200-Fig05-For%20the%20simple%20model%20outlined%20in%20the%20text,%20whether%20one%20should%20focus%20attention%20on%20only%20the%20highest%20valued%20action%20or%20divide%20attention%20between%20the%20two%20best%20actions,%20as%20a....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=722>) · Original paper and literature note are not hosted.
### P202 - Fig. 1
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 1 | Triadic model of SRL. Adapted from Zimmerman (1989).
**Purpose:** Original SRL theoretical framework reproduced for conceptual comparison; full figure and attribution retained. **Visual forms:** Conceptual diagram.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P202-Fig01-Triadic%20model%20of%20SRL.%20Adapted%20from%20Zimmerman%20(1989).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=723>) · Original paper and literature note are not hosted.
### P202 - Fig. 2
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 2 | Cyclical phases model (1st version). Adapted from Zimmerman (2000).
**Purpose:** Figure 2 comprises both the cyclical diagram and phase/subprocess table in one original framed figure. **Visual forms:** Conceptual diagram, Complex table.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P202-Fig02-Cyclical%20phases%20model%20(1st%20version).%20Adapted%20from%20Zimmerman%20(2000).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=724>) · Original paper and literature note are not hosted.
### P202 - Table 2
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** TABLE 2 | Models’ phases. *The early draft provided by the authors did not provide the exact names for the phases but it could be implied the phases are similar to Winne and Hadwin’s. Therefore, this review comparison will be based on their 2011 publication.
**Purpose:** Structured theoretical comparison aligns seven model/version rows across preparatory, performance and appraisal phases; full historical-version note retained. **Visual forms:** Complex table.
**Physical source pages:** 19.
[PNG](<01_Theory_and_Hypotheses/P202-Table02-Models%E2%80%99%20phases.%201%20asterisk%20The%20early%20draft%20provided%20by%20the%20authors%20did%20not%20provide%20the%20exact%20names%20for%20the%20phases%20but%20it%20could%20be%20implied%20the%20phases%20are%20similar%20to....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=725>) · Original paper and literature note are not hosted.
### P202 - Fig. 3
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 3 | Current version Cyclical phases model. Adapted from Zimmerman and Moylan (2009).
**Purpose:** Zimmerman cyclical-phase theoretical framework; full source attribution restored. **Visual forms:** Conceptual diagram.
**Physical source pages:** 5.
[PNG](<01_Theory_and_Hypotheses/P202-Fig03-Current%20version%20Cyclical%20phases%20model.%20Adapted%20from%20Zimmerman%20and%20Moylan%20(2009).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=726>) · Original paper and literature note are not hosted.
### P202 - Table 3
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** TABLE 3 | Models figures comparison on cognition, motivation, and emotion. Note from the author: these levels should be read as a continuum.
**Purpose:** Distinctive qualitative matrix positions models on relative theoretical emphases, with interpretive note. **Visual forms:** Qualitative comparison matrix.
**Physical source pages:** 19.
[PNG](<01_Theory_and_Hypotheses/P202-Table03-Models%20figures%20comparison%20on%20cognition,%20motivation,%20and%20emotion.%20Note%20from%20the%20author%20these%20levels%20should%20be%20read%20as%20a%20continuum.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=727>) · Original paper and literature note are not hosted.
### P202 - Fig. 4
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 4 | Multi-level model. Adapted from Zimmerman (2000).
**Purpose:** Original SRL theoretical framework reproduced for conceptual comparison; full figure and attribution retained. **Visual forms:** Conceptual diagram.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P202-Fig04-Multi-level%20model.%20Adapted%20from%20Zimmerman%20(2000).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=728>) · Original paper and literature note are not hosted.
### P202 - Fig. 5
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 5 | Six-component model of SRL. Adapted from Boekaerts (1996b).
**Purpose:** Boekaerts six-component theoretical framework; full framed figure, caption and attribution retained. **Visual forms:** Conceptual diagram.
**Physical source pages:** 7.
[PNG](<01_Theory_and_Hypotheses/P202-Fig05-Six-component%20model%20of%20SRL.%20Adapted%20from%20Boekaerts%20(1996b).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=729>) · Original paper and literature note are not hosted.
### P202 - Fig. 6
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 6 | Model of adaptable learning. Extracted from the 2000 handbook but cited there as: the original model of adaptable learning. Adapted from Boekaerts (1996a).
**Purpose:** Original SRL theoretical framework reproduced for conceptual comparison; full figure and attribution retained. **Visual forms:** Conceptual diagram.
**Physical source pages:** 8.
[PNG](<01_Theory_and_Hypotheses/P202-Fig06-Model%20of%20adaptable%20learning.%20Extracted%20from%20the%202000%20handbook%20but%20cited%20there%20as%20the%20original%20model%20of%20adaptable%20learning.%20Adapted%20from%20Boekaerts%20(1996a).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=730>) · Original paper and literature note are not hosted.
### P202 - Fig. 7
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 7 | Dual processing self-regulation model. Adapted from Boekaerts (2011).
**Purpose:** Original SRL theoretical framework reproduced for conceptual comparison; full figure and attribution retained. **Visual forms:** Conceptual diagram.
**Physical source pages:** 9.
[PNG](<01_Theory_and_Hypotheses/P202-Fig07-Dual%20processing%20self-regulation%20model.%20Adapted%20from%20Boekaerts%20(2011).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=731>) · Original paper and literature note are not hosted.
### P202 - Fig. 8
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 8 | First version of Winne’s SRL model. Adapted from Winne (1996).
**Purpose:** Original SRL theoretical framework reproduced for conceptual comparison; full figure and attribution retained. **Visual forms:** Conceptual diagram.
**Physical source pages:** 11.
[PNG](<01_Theory_and_Hypotheses/P202-Fig08-First%20version%20of%20Winne%E2%80%99s%20SRL%20model.%20Adapted%20from%20Winne%20(1996).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=732>) · Original paper and literature note are not hosted.
### P202 - Fig. 9
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 9 | Current version of Winne’s SRL model. Adapted from Winne and Hadwin (1998).
**Purpose:** Winne/Hadwin COPES theoretical architecture; original complete caption restored. **Visual forms:** Conceptual diagram.
**Physical source pages:** 12.
[PNG](<01_Theory_and_Hypotheses/P202-Fig09-Current%20version%20of%20Winne%E2%80%99s%20SRL%20model.%20Adapted%20from%20Winne%20and%20Hadwin%20(1998).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=733>) · Original paper and literature note are not hosted.
### P202 - Fig. 10
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 10 | Pintrich’s SRL model. Adapted from Pintrich (2000).
**Purpose:** Original conceptual phase-by-regulation-area figure; separate from adjacent Figure 11. **Visual forms:** Conceptual matrix.
**Physical source pages:** 14.
[PNG](<01_Theory_and_Hypotheses/P202-Fig10-Pintrich%E2%80%99s%20SRL%20model.%20Adapted%20from%20Pintrich%20(2000).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=734>) · Original paper and literature note are not hosted.
### P202 - Fig. 11
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 11 | Metacognitive and affective Model of Self-Regulated Learning model (MASRL). Adapted from Efklides (2011).
**Purpose:** Distinct original Efklides model had been merged into Figure 10 detection; both person and task-by-person levels and complete attribution retained. **Visual forms:** Conceptual diagram, Composite.
**Physical source pages:** 14.
[PNG](<01_Theory_and_Hypotheses/P202-Fig11-Metacognitive%20and%20affective%20Model%20of%20Self-Regulated%20Learning%20model%20(MASRL).%20Adapted%20from%20Efklides%20(2011).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=735>) · Original paper and literature note are not hosted.
### P202 - Fig. 12
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 12 | Socially shared regulated learning model 1. Adapted from Järvelä and Hadwin (2013).
**Purpose:** Original SRL theoretical framework reproduced for conceptual comparison; full figure and attribution retained. **Visual forms:** Conceptual diagram.
**Physical source pages:** 17.
[PNG](<01_Theory_and_Hypotheses/P202-Fig12-Socially%20shared%20regulated%20learning%20model%201.%20Adapted%20from%20J%C3%A4rvel%C3%A4%20and%20Hadwin%20(2013).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=736>) · Original paper and literature note are not hosted.
### P202 - Fig. 13
**Paper:** A Review of Self-regulated Learning: Six Models and Four Directions for Research
**Original caption:** FIGURE 13 | Socially shared regulated learning model 2. Adapted from Hadwin et al. (2011).
**Purpose:** Original model of shared regulation; full attribution restored. **Visual forms:** Conceptual diagram.
**Physical source pages:** 18.
[PNG](<01_Theory_and_Hypotheses/P202-Fig13-Socially%20shared%20regulated%20learning%20model%202.%20Adapted%20from%20Hadwin%20et%20al.%20(2011).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=737>) · Original paper and literature note are not hosted.
### P210 - Fig. 3
**Paper:** Cognitive Offloading
**Original caption:** Figure 3 Caption: A metacognitive model of cognitive offloading. We propose that selecting between offloading and relying on internal processes is influenced by metacognitive evaluations of our (internal) mental capacities and the capacities of our extended mental systems encompassing body and world (arrow a). An example of this would be evaluating our unaided spatial memory and a GPS system when deciding how to navigate to a friend’s house. In addition, engaging in either internal or extended strategies can influence subsequent metacognitive evaluations (arrow b). For instance, after successful use of a GPS system we may come to believe that it is a more reliable guide than our unaided memory. Offloading can also directly impact our lower-level cognitive processes (arrow c). An example of this would be a reduction in our internal spatial memory for a location after reliance on GPS navigation.
**Purpose:** Proposed metacognitive feedback framework, not a tested coefficient model; detached original caption retained. **Visual forms:** Conceptual diagram.
**Physical source pages:** 29, 37.
[PNG](<01_Theory_and_Hypotheses/P210-Fig03-Caption%20A%20metacognitive%20model%20of%20cognitive%20offloading.%20We%20propose%20that%20selecting%20between%20offloading%20and%20relying%20on%20internal%20processes%20is%20influenced%20by%20metacognitive....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=738>) · Original paper and literature note are not hosted.
### P219 - Fig. 1
**Paper:** Explaining social disparities in ICT literacy: the role of multiple socioeconomic status indicators and ICT-related motivations
**Original caption:** Fig. 1 Illustration of the mediation models (a: cross-sectional, b: longitudinal). For ease of presentation, no differentiation was made between the SES indicators (cultural capital, parents` educational level and occupational status). Correlations between the independent variables (SES indicators, intelligence, ICT Literacy Grade 9) and between the mediator variables (instrumental and social-interactive ICT usage motives, reading behavior) are not shown (see Table 1)
**Purpose:** Hypothesized cross-sectional and longitudinal mediation frameworks, before coefficient estimation. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 10.
[PNG](<01_Theory_and_Hypotheses/P219-Fig01-Illustration%20of%20the%20mediation%20models%20(a%20cross-sectional,%20b%20longitudinal).%20For%20ease%20of%20presentation,%20no%20differentiation%20was%20made%20between%20the%20SES%20indicators%20(cultural....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=739>) · Original paper and literature note are not hosted.
### P222 - Fig. 2
**Paper:** From byproduct to design factor: on validating the interpretation of process indicators based on log data
**Original caption:** Fig. 2 Assessment as reasoning from evidence derived from log data
**Purpose:** Conceptual framework linking evidence-centered design to log-based construct inference. **Visual forms:** Flowchart.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P222-Fig02-Assessment%20as%20reasoning%20from%20evidence%20derived%20from%20log%20data.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=740>) · Original paper and literature note are not hosted.
### P228 - Table 1
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Table 1 Categories and subcategories of the taxonomy of paradata with prototypical examples
**Purpose:** Structured conceptual taxonomy linking categories, definitions and observation levels. **Visual forms:** Complex table.
**Physical source pages:** 5.
[PNG](<01_Theory_and_Hypotheses/P228-Table01-Categories%20and%20subcategories%20of%20the%20taxonomy%20of%20paradata%20with%20prototypical%20examples.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=741>) · Original paper and literature note are not hosted.
### P228 - Fig. 1
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Fig. 1 Taxonomy of paradata with categories access-related, response-related and process-related paradata
**Purpose:** Conceptual taxonomy. **Visual forms:** Composite.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P228-Fig01-Taxonomy%20of%20paradata%20with%20categories%20access-related,%20response-related%20and%20process-related%20paradata.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=742>) · Original paper and literature note are not hosted.
### P231 - Fig. 5
**Paper:** Inequality of opportunity in non-cognitive skills: a cross-region analysis
**Original caption:** Fig. 5 Roemer’s theoretical framework.
**Purpose:** Roemer framework separates direct and indirect circumstance effects from effort-based inequality. **Visual forms:** Path diagram.
**Physical source pages:** 11.
[PNG](<01_Theory_and_Hypotheses/P231-Fig05-Roemer%E2%80%99s%20theoretical%20framework.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=743>) · Original paper and literature note are not hosted.
### P234 - Fig. 1
**Paper:** Large studies reveal how reference bias limits policy applications of self-report measures
**Original caption:** Figure 1. Peers influence the standards by which an individual judges their own behavior, resulting in a “reference bias” effect that distorts cross-context comparisons of self-reported self-regulation. Illustration by Tom McQuaid.
**Purpose:** Illustrates how peer reference standards alter self-evaluation. **Visual forms:** Composite.
**Physical source pages:** 2.
[PNG](<01_Theory_and_Hypotheses/P234-Fig01-Peers%20influence%20the%20standards%20by%20which%20an%20individual%20judges%20their%20own%20behavior,%20resulting%20in%20a%20%E2%80%9Creference%20bias%E2%80%9D%20effect%20that%20distorts%20cross-context%20comparisons%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=744>) · Original paper and literature note are not hosted.
### P242 - Fig. 1
**Paper:** Self-Regulated Learning: Beliefs, Techniques, and Illusions
**Original caption:** Figure 1. Adapted from Nelson & Narens’s (1990) framework for metamemory. From Dunlosky et al. (2007).
**Purpose:** Metamemory monitoring/control framework; full diagram and attribution retained. **Visual forms:** Conceptual diagram.
**Physical source pages:** 6.
[PNG](<01_Theory_and_Hypotheses/P242-Fig01-Adapted%20from%20Nelson%20&%20Narens%E2%80%99s%20(1990)%20framework%20for%20metamemory.%20From%20Dunlosky%20et%20al.%20(2007).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=745>) · Original paper and literature note are not hosted.
### P242 - Fig. 3
**Paper:** Self-Regulated Learning: Beliefs, Techniques, and Illusions
**Original caption:** Figure 3. Hypothetical curves depicting changes in predicted and actual learning over time. Predicted recall remains constant due to the stability bias, whereas actual recall increases due to future study trials (left panel) or decreases due to forgetting (right panel). These dynamics produce an increasing gap between predicted and actual knowledge.
**Purpose:** Hypothetical illustration of stability bias, not observed empirical growth curves. **Visual forms:** Conceptual curves, Composite.
**Physical source pages:** 13.
[PNG](<01_Theory_and_Hypotheses/P242-Fig03-Hypothetical%20curves%20depicting%20changes%20in%20predicted%20and%20actual%20learning%20over%20time.%20Predicted%20recall%20remains%20constant%20due%20to%20the%20stability%20bias,%20whereas%20actual%20recall....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=746>) · Original paper and literature note are not hosted.
### P249 - Table 1
**Paper:** The Effort Paradox: Effort Is Both Costly and Valued
**Original caption:** Table 1. Effort As Value: Theories and Phenomena
**Purpose:** Structured comparison of effort-valuation theories, mechanisms, targets and timing; a distinctive conceptual evidence table. **Visual forms:** Complex table, Taxonomy.
**Physical source pages:** 4, 5.
[PNG](<01_Theory_and_Hypotheses/P249-Table01-Effort%20As%20Value%20Theories%20and%20Phenomena.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=747>) · Original paper and literature note are not hosted.
### P249 - Fig. I
**Paper:** The Effort Paradox: Effort Is Both Costly and Valued
**Original caption:** Figure I. Two Alternative Formulations for the Positive Valuation of Effort. Previous work has suggested that the expected value of effort (EVprev ; dark blue lines) is the difference between the expected product of investing that amount of effort (i.e., the likelihood of reaching your goal multiplied by the incentives received when the goal is reached; dark green lines) and the cost of exerting that same intensity of effort (red lines). The individual can (in principle) select their optimal effort investment by locating the effort intensity that maximizes their EV (unbroken vertical arrow). This framework can accommodate two alternate ways of valuing effort (shown in green broken lines): (A) as an amplification in the value of the product of effort with increasing effort intensity (from dark green to light green values), and/or (B) as a positive value that increases with effort intensity (counteracting effort costs), separate from the positive value expected as the product of effort. Effort’s reward function is assumed to be linear in this example, but the actual forms of this function and the cost function remain to be determined. Each of these alterations results in an alternate set of EV values (EValt ; broken blue lines) and corresponding changes in the optimal level of effort.
**Purpose:** Two-panel theoretical comparison of product-centered and effort-centered reward formulations. **Visual forms:** Conceptual curve, Composite.
**Physical source pages:** 12.
[PNG](<01_Theory_and_Hypotheses/P249-FigI-Two%20Alternative%20Formulations%20for%20the%20Positive%20Valuation%20of%20Effort.%20Previous%20work%20has%20suggested%20that%20the%20expected%20value%20of%20effort%20(EVprev%20;%20dark%20blue%20lines)%20is%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=748>) · Original paper and literature note are not hosted.
### P251 - Fig. 1
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 1. Illustration of the Brunswik symmetry principle according to Wittmann [23,24]. Dotted lines = symmetrical comparison. Dashed lines = asymmetrical comparison.
**Purpose:** Explains construct hierarchies and symmetry. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 3.
[PNG](<01_Theory_and_Hypotheses/P251-Fig01-Illustration%20of%20the%20Brunswik%20symmetry%20principle%20according%20to%20Wittmann%20%5B23,24%5D.%20Dotted%20lines%20=%20symmetrical%20comparison.%20Dashed%20lines%20=%20asymmetrical%20comparison.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=749>) · Original paper and literature note are not hosted.
### P251 - Fig. 2
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 2. Illustration of the hierarchical levels. GFP = general factor of personality, C = Conscientiousness, A = Agreeableness, ES = Emotional Stability, O = Openness, E = Extraversion, Numbers = facets of the corresponding dimensions, Rea = reasoning, Flu = fluency, Spe = perceptual speed, Mem = short-term memory, ver = verbal, num = numerical, fig = figural. The lowest level (i.e., items and specific responses) of the Pyramidal Model of Personality [4,5] are not displayed.
**Purpose:** Explains construct hierarchies and symmetry. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 4.
[PNG](<01_Theory_and_Hypotheses/P251-Fig02-Illustration%20of%20the%20hierarchical%20levels.%20GFP%20=%20general%20factor%20of%20personality,%20C%20=%20Conscientiousness,%20A%20=%20Agreeableness,%20ES%20=%20Emotional%20Stability,%20O%20=%20Openness,%20E%20=....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=750>) · Original paper and literature note are not hosted.
## 02 Study Design and Sampling
### P058 - Fig. 3
**Paper:** Can inquiry-based pedagogy improve math performance? Evidence from 5711 students in Vietnam on the mediating role of math attitude
**Original caption:** Fig. 3. : Flowchart of subject inclusion.
**Purpose:** Respondent inclusion and missing-data exclusions. **Visual forms:** Flowchart.
**Physical source pages:** 3.
[PNG](<02_Study_Design_and_Sampling/P058-Fig03-Flowchart%20of%20subject%20inclusion.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=62>) · Original paper and literature note are not hosted.
### P065 - Fig. 2
**Paper:** The impact of CT-integrated mathematics instruction on secondary school students' computational thinking skills
**Original caption:** Fig. 2. Experimental research idea diagram
**Purpose:** Experimental/control sequence and measurement timeline. **Visual forms:** Flowchart.
**Physical source pages:** 4.
[PNG](<02_Study_Design_and_Sampling/P065-Fig02-Experimental%20research%20idea%20diagram.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=63>) · Original paper and literature note are not hosted.
### P082 - Fig. 1
**Paper:** National contexts and information and communication technology in education: A systematic review of PISA studies
**Original caption:** Fig. 1. The Search and Screening Processes This figure illustrates the PRISMA-based flow of article identification, screening, eligibility checks, and inclusion for the systematic review.
**Purpose:** PRISMA selection flow for the review corpus. **Visual forms:** Flowchart.
**Physical source pages:** 4.
[PNG](<02_Study_Design_and_Sampling/P082-Fig01-The%20Search%20and%20Screening%20Processes%20This%20figure%20illustrates%20the%20PRISMA-based%20flow%20of%20article%20identification,%20screening,%20eligibility%20checks,%20and%20inclusion%20for%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=64>) · Original paper and literature note are not hosted.
### P090 - Fig. 2
**Paper:** Meta-affective progression in science: A cross-sectional study of junior high school students
**Original caption:** Fig. 2. Two-group cross-sectional design with meta-affective variables.
**Purpose:** Two-grade cross-sectional comparative study design. **Visual forms:** Flowchart.
**Physical source pages:** 4.
[PNG](<02_Study_Design_and_Sampling/P090-Fig02-Two-group%20cross-sectional%20design%20with%20meta-affective%20variables.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=65>) · Original paper and literature note are not hosted.
### P094 - Table 1
**Paper:** What teachers can do to enhance students’ well-being: Discussion
**Original caption:** Table 1. Topics and samples.
**Purpose:** Compares topics and sample characteristics of the included studies. **Visual forms:** Complex table, Structured evidence comparison.
**Physical source pages:** 2.
[PNG](<02_Study_Design_and_Sampling/P094-Table01-Topics%20and%20samples.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=66>) · Original paper and literature note are not hosted.
### P107 - Fig. 1
**Paper:** The association between teacher self-efficacy and student academic performance: A systematic review and meta-analysis
**Original caption:** Fig. 1. Flow diagram of literature search.
**Purpose:** Documents literature screening and inclusion. **Visual forms:** PRISMA flowchart.
**Physical source pages:** 7.
[PNG](<02_Study_Design_and_Sampling/P107-Fig01-Flow%20diagram%20of%20literature%20search.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=67>) · Original paper and literature note are not hosted.
### P118 - Fig. 2
**Paper:** Understanding the relationship between teacher collaboration beliefs and job satisfaction via collaborative attitudes: A moderated mediation model of classroom self-efficacy from PISA 2022
**Original caption:** Fig. 2. Excluded and included number of observations.
**Purpose:** Sample exclusions and final included teacher observations. **Visual forms:** Flowchart.
**Physical source pages:** 6.
[PNG](<02_Study_Design_and_Sampling/P118-Fig02-Excluded%20and%20included%20number%20of%20observations.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=68>) · Original paper and literature note are not hosted.
### P119 - Table 1
**Paper:** Exploring Initial Teacher Education students’ learning journeys while implementing a reading intervention program
**Original caption:** Table 1 Research project data collection details.
**Purpose:** Compares data-collection methods, participants, collected evidence and timing. **Visual forms:** Complex table.
**Physical source pages:** 3.
[PNG](<02_Study_Design_and_Sampling/P119-Table01-Research%20project%20data%20collection%20details.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=69>) · Original paper and literature note are not hosted.
### P128 - Fig. 1
**Paper:** “Together we are better”: Professional learning networks for teachers
**Original caption:** Fig. 1. Screenshot of request for participation tweet.
**Purpose:** Illustrates the social-media recruitment material used to obtain the sample. **Visual forms:** Interface.
**Physical source pages:** 4.
[PNG](<02_Study_Design_and_Sampling/P128-Fig01-Screenshot%20of%20request%20for%20participation%20tweet.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=70>) · Original paper and literature note are not hosted.
### P129 - Fig. 2
**Paper:** A bibliometric and thematic analysis of PISA literature for school improvement and effectiveness over two decades
**Original caption:** Fig. 2. Flow chart of search and screen process.
**Purpose:** Literature search and eligibility screening. **Visual forms:** Flowchart.
**Physical source pages:** 4.
[PNG](<02_Study_Design_and_Sampling/P129-Fig02-Flow%20chart%20of%20search%20and%20screen%20process.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=71>) · Original paper and literature note are not hosted.
### P135 - Table 1
**Paper:** The application of machine learning approaches in international large-scale educational assessments: A scoping review
**Original caption:** Table 1 Search strategy.
**Purpose:** Compares database-specific search strings and restrictions used to form the review sample. **Visual forms:** Complex table.
**Physical source pages:** 4.
[PNG](<02_Study_Design_and_Sampling/P135-Table01-Search%20strategy.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=72>) · Original paper and literature note are not hosted.
### P135 - Fig. 2
**Paper:** The application of machine learning approaches in international large-scale educational assessments: A scoping review
**Original caption:** Fig. 2. The frequency of machine learning approaches. Note. For the purpose of synthesis, we grouped similar machine learning approaches into broader categories. For example, XGBoost, LightGBM, and Adaptive Boosting algorithms were all classified under the umbrella category of Boosting algorithms. A detailed list of the specific approaches employed in each study can be found in Table S1 of the Supplementary Material.
**Purpose:** PRISMA-style search and inclusion flow; source caption mismatch retained. **Visual forms:** Flowchart.
**Physical source pages:** 6.
[PNG](<02_Study_Design_and_Sampling/P135-Fig02-The%20frequency%20of%20machine%20learning%20approaches.%20Note.%20For%20the%20purpose%20of%20synthesis,%20we%20grouped%20similar%20machine%20learning%20approaches%20into%20broader%20categories.%20For....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=73>) · Original paper and literature note are not hosted.
### P146 - Table 1
**Paper:** A meta-analysis on the effect of technology on the achievement of less advantaged students
**Original caption:** Table 1 Inclusion and exclusion criteria.
**Purpose:** Systematic-review eligibility rules align inclusion and exclusion logic. **Visual forms:** Complex table.
**Physical source pages:** 4.
[PNG](<02_Study_Design_and_Sampling/P146-Table01-Inclusion%20and%20exclusion%20criteria.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=74>) · Original paper and literature note are not hosted.
### P146 - Fig. 1
**Paper:** A meta-analysis on the effect of technology on the achievement of less advantaged students
**Original caption:** Fig. 1. Flowchart illustrating the review selection process.
**Purpose:** Review screening and full-text eligibility process. **Visual forms:** Flowchart.
**Physical source pages:** 6.
[PNG](<02_Study_Design_and_Sampling/P146-Fig01-Flowchart%20illustrating%20the%20review%20selection%20process.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=75>) · Original paper and literature note are not hosted.
### P149 - Fig. 1
**Paper:** A systematic review and meta-analysis of the evidence on learning during the COVID-19 pandemic
**Original caption:** Fig. 1 | Flow diagram of the study identification and selection process, following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.
**Purpose:** Documents study identification, eligibility and inclusion. **Visual forms:** PRISMA flowchart.
**Physical source pages:** 2.
[PNG](<02_Study_Design_and_Sampling/P149-Fig01-Flow%20diagram%20of%20the%20study%20identification%20and%20selection%20process,%20following%20Preferred%20Reporting%20Items%20for%20Systematic%20Reviews%20and%20Meta-Analyses%20(PRISMA)%20guidelines.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=76>) · Original paper and literature note are not hosted.
### P156 - Fig. 1
**Paper:** Effects of LLM use and note-taking on reading comprehension and memory: A randomised experiment in secondary schools
**Original caption:** Fig. 1. Study design illustrating the activities and their order during Session 1 and 2.
**Purpose:** Randomized session sequence and task allocation. **Visual forms:** Flowchart, Study timeline.
**Physical source pages:** 6.
[PNG](<02_Study_Design_and_Sampling/P156-Fig01-Study%20design%20illustrating%20the%20activities%20and%20their%20order%20during%20Session%201%20and%202.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=77>) · Original paper and literature note are not hosted.
### P158 - Fig. 3
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 3 Search string based on the PICO framework
**Purpose:** PICO search-string construction defines evidence identification. **Visual forms:** Complex table, Search strategy.
**Physical source pages:** 12.
[PNG](<02_Study_Design_and_Sampling/P158-Fig03-Search%20string%20based%20on%20the%20PICO%20framework.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=78>) · Original paper and literature note are not hosted.
### P158 - Fig. 4
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 4 PRISMA flow diagram of study selection for review and meta-analysis
**Purpose:** Evidence screening and inclusion. **Visual forms:** PRISMA flowchart.
**Physical source pages:** 20.
[PNG](<02_Study_Design_and_Sampling/P158-Fig04-PRISMA%20flow%20diagram%20of%20study%20selection%20for%20review%20and%20meta-analysis.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=79>) · Original paper and literature note are not hosted.
### P158 - Table 4
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Table 4 Studies included in the review. p < 0.05 = *; p < 0.01 = **; p < 0.001 = ***. 1 GAI vs. GAI
**Purpose:** Multi-level partitions compare study designs, samples, AI systems and outcome evidence. **Visual forms:** Complex table, Structured evidence comparison.
**Physical source pages:** 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85.
[PNG part 1](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part01.png>) · [PNG part 2](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part02.png>) · [PNG part 3](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part03.png>) · [PNG part 4](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part04.png>) · [PNG part 5](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part05.png>) · [PNG part 6](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part06.png>) · [PNG part 7](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part07.png>) · [PNG part 8](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part08.png>) · [PNG part 9](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part09.png>) · [PNG part 10](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part10.png>) · [PNG part 11](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part11.png>) · [PNG part 12](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part12.png>) · [PNG part 13](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part13.png>) · [PNG part 14](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part14.png>) · [PNG part 15](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part15.png>) · [PNG part 16](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part16.png>) · [PNG part 17](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part17.png>) · [PNG part 18](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part18.png>) · [PNG part 19](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part19.png>) · [PNG part 20](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part20.png>) · [PNG part 21](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part21.png>) · [PNG part 22](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part22.png>) · [PNG part 23](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part23.png>) · [PNG part 24](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part24.png>) · [PNG part 25](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part25.png>) · [PNG part 26](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part26.png>) · [PNG part 27](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part27.png>) · [PNG part 28](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part28.png>) · [PNG part 29](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part29.png>) · [PNG part 30](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part30.png>) · [PNG part 31](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part31.png>) · [PNG part 32](<02_Study_Design_and_Sampling/P158-Table04-Studies%20included%20in%20the%20review.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks%20.%201%20GAI%20vs.%20GAI-Part32.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=80>) · Original paper and literature note are not hosted.
### P158 - Fig. 13
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 13 Detailed PRISMA flow diagram of study selection for review and meta-analysis detailed for forwards and backwards citation
**Purpose:** Detailed forward/backward citation screening. **Visual forms:** PRISMA flowchart.
**Physical source pages:** 53.
[PNG](<02_Study_Design_and_Sampling/P158-Fig13-Detailed%20PRISMA%20flow%20diagram%20of%20study%20selection%20for%20review%20and%20meta-analysis%20detailed%20for%20forwards%20and%20backwards%20citation.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=112>) · Original paper and literature note are not hosted.
### P159 - Fig. 1
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure 1: Experimental Session Timelines. Panel A. Session One timeline. Panel B. Session Two timeline. Notes: This figure shows the timeline of the two experimental sessions. Panel A shows the structure of Session One, during which students were randomly assigned to AI-allowed or AI-forbidden conditions. Panel B shows Session Two, conducted approximately one week later (mean of 6.98 days), during which all students completed tasks without AI access. The red dashed box in Panel B indicates that the order of the Endline Test and the Essay was randomized across students. The Baseline Test and Post-Learning Test in Session One each contained five multiple-choice questions. The Endline Test in Session Two contained ten multiple-choice questions.
**Purpose:** Shows the sequence and timing of the two randomized experimental sessions. **Visual forms:** Line/profile plot.
**Physical source pages:** 28.
[PNG](<02_Study_Design_and_Sampling/P159-Fig01-Experimental%20Session%20Timelines.%20Panel%20A.%20Session%20One%20timeline.%20Panel%20B.%20Session%20Two%20timeline.%20Notes%20This%20figure%20shows%20the%20timeline%20of%20the%20two%20experimental%20sessions....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=113>) · Original paper and literature note are not hosted.
### P159 - Fig. A1
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure A1: Computer Lab Setup with Privacy Dividers. Notes: This figure shows the computer lab setup used during experimental sessions. Each workstation was equipped with privacy dividers to minimize distractions and prevent participants from viewing other screens.
**Purpose:** Documents the laboratory setup and staff procedures used to run and monitor the experiment. **Visual forms:** Composite.
**Physical source pages:** 43.
[PNG](<02_Study_Design_and_Sampling/P159-FigA1-Computer%20Lab%20Setup%20with%20Privacy%20Dividers.%20Notes%20This%20figure%20shows%20the%20computer%20lab%20setup%20used%20during%20experimental%20sessions.%20Each%20workstation%20was%20equipped%20with....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=114>) · Original paper and literature note are not hosted.
### P159 - Fig. A3
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure A3: Laboratory Seating Charts. Panel A. Library computer lab (LIB 140). Panel B. Language Building computer lab (SDL 122). Notes: This figure shows the seating charts used by laboratory staff to monitor compliance during experimental sessions. Each numbered square represents a computer workstation. For each session time slot, one lab was randomly assigned to the AI-allowed condition and the other to the AI-forbidden condition. Proctors used these charts to record any instances of unauthorized resource use, with two proctors assigned to each laboratory. The physical separation of treatment conditions across different buildings helped prevent cross-contamination between experimental groups.
**Purpose:** Documents the laboratory setup and staff procedures used to run and monitor the experiment. **Visual forms:** Composite.
**Physical source pages:** 45.
[PNG](<02_Study_Design_and_Sampling/P159-FigA3-Laboratory%20Seating%20Charts.%20Panel%20A.%20Library%20computer%20lab%20(LIB%20140).%20Panel%20B.%20Language%20Building%20computer%20lab%20(SDL%20122).%20Notes%20This%20figure%20shows%20the%20seating%20charts....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=115>) · Original paper and literature note are not hosted.
### P183 - Fig. 3
**Paper:** PISA 2018 in England, Northern Ireland, Scotland and Wales: Is the data really representative of all four corners of the UK?
**Original caption:** Figure 3. Illustration of key features of the PISA sample design.
**Purpose:** Maps the PISA sample construction stages to technical standards. **Visual forms:** Line/profile plot.
**Physical source pages:** 10.
[PNG](<02_Study_Design_and_Sampling/P183-Fig03-Illustration%20of%20key%20features%20of%20the%20PISA%20sample%20design.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=116>) · Original paper and literature note are not hosted.
### P183 - Table A2
**Paper:** PISA 2018 in England, Northern Ireland, Scotland and Wales: Is the data really representative of all four corners of the UK?
**Original caption:** Table A2. England.
**Purpose:** Multi-panel comparison of original, replacement, and weighted school samples. **Visual forms:** Complex table.
**Physical source pages:** 40.
[PNG](<02_Study_Design_and_Sampling/P183-TableA2-England.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=117>) · Original paper and literature note are not hosted.
### P183 - Table A3
**Paper:** PISA 2018 in England, Northern Ireland, Scotland and Wales: Is the data really representative of all four corners of the UK?
**Original caption:** Table A3. Northern Ireland.
**Purpose:** Multi-panel comparison of original, replacement, and weighted school samples. **Visual forms:** Complex table.
**Physical source pages:** 40.
[PNG](<02_Study_Design_and_Sampling/P183-TableA3-Northern%20Ireland.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=118>) · Original paper and literature note are not hosted.
### P188 - Fig. 2
**Paper:** Socioeconomic status and student learning: insights from an umbrella review
**Original caption:** Fig. 2 PRISMA
**Purpose:** PRISMA review-selection flow. **Visual forms:** Flowchart.
**Physical source pages:** 14.
[PNG](<02_Study_Design_and_Sampling/P188-Fig02-PRISMA.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=119>) · Original paper and literature note are not hosted.
### P192 - Table 1
**Paper:** Supporting Self-Regulated Learning with Generative AI: A Case of Two Empirical Studies
**Original caption:** Table 1 Overview of the Four Experimental conditions
**Purpose:** Parallel experimental-condition timeline compares four planned summary/GenAI interventions and common measurements. **Visual forms:** Complex table.
**Physical source pages:** 4, 5.
[PNG](<02_Study_Design_and_Sampling/P192-Table01-Overview%20of%20the%20Four%20Experimental%20conditions.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=120>) · Original paper and literature note are not hosted.
### P203 - Fig. 1
**Paper:** A Systematic Review and Meta-Analysis of Discrepancies Between Logged and Self-Reported Digital Media Use
**Original caption:** Figure 1. PRISMA flow diagram for the study inclusion process. A total of 47 records fulfilled the eligibility criteria.
**Purpose:** PRISMA study selection and screening. **Visual forms:** Flowchart.
**Physical source pages:** 26, 38.
[PNG](<02_Study_Design_and_Sampling/P203-Fig01-PRISMA%20flow%20diagram%20for%20the%20study%20inclusion%20process.%20A%20total%20of%2047%20records%20fulfilled%20the%20eligibility%20criteria.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=751>) · Original paper and literature note are not hosted.
### P205 - Fig. 1
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** FIGURE 1 Experimental procedure
**Purpose:** Randomized four-agent experimental timeline links stages, measures and research questions; caption restored below original automatic crop. **Visual forms:** Study timeline, Composite.
**Physical source pages:** 8.
[PNG](<02_Study_Design_and_Sampling/P205-Fig01-Experimental%20procedure.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=752>) · Original paper and literature note are not hosted.
### P206 - Fig. 1
**Paper:** Beyond traditional interviews: Psychometric analysis of asynchronous video interviews for personality and interview performance evaluation using machine learning
**Original caption:** Fig. 1. Flowchart with inclusion and exclusion criteria for eligible participants.
**Purpose:** Caption/context matches prisma|flow\s*chart|flow diagram|screening|sample selection|sampling|study design|participant.*flow|literature search. **Visual forms:** Flowchart.
**Physical source pages:** 6.
[PNG](<02_Study_Design_and_Sampling/P206-Fig01-Flowchart%20with%20inclusion%20and%20exclusion%20criteria%20for%20eligible%20participants.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=753>) · Original paper and literature note are not hosted.
### P211 - Fig. 1
**Paper:** Comparisons of statistical methods for handling attrition in a follow-up visit with complex survey sampling
**Original caption:** FIGURE 1 Flowchart to illustrate sampling design for simulated data
**Purpose:** Three-stage stratified sampling design used to generate the simulations; all strata, sampling probabilities and caption retained. **Visual forms:** Flowchart.
**Physical source pages:** 9.
[PNG](<02_Study_Design_and_Sampling/P211-Fig01-Flowchart%20to%20illustrate%20sampling%20design%20for%20simulated%20data.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=754>) · Original paper and literature note are not hosted.
### P215 - Fig. 1
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Fig. 1. Article search and selection process.
**Purpose:** Review identification, screening and eligibility process, not bibliometric mapping. **Visual forms:** PRISMA flowchart.
**Physical source pages:** 8.
[PNG](<02_Study_Design_and_Sampling/P215-Fig01-Article%20search%20and%20selection%20process.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=755>) · Original paper and literature note are not hosted.
### P253 - Table 1
**Paper:** Validity and fairness of the PISA 2018 Global Competence assessment: an argument-based evaluation via explanatory item response models
**Original caption:** Table 1 Presence of testlets and booklets across booklet groups.
**Purpose:** Grouped mapping of booklet forms, overlapping testlet sets and sample counts documents the matrix-sampled assessment design, beyond an ordinary summary table. **Visual forms:** Complex table, Design mapping.
**Physical source pages:** 5.
[PNG](<02_Study_Design_and_Sampling/P253-Table01-Presence%20of%20testlets%20and%20booklets%20across%20booklet%20groups.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=756>) · Original paper and literature note are not hosted.
## 03 Data Description
### P058 - Fig. 2
**Paper:** Can inquiry-based pedagogy improve math performance? Evidence from 5711 students in Vietnam on the mediating role of math attitude
**Original caption:** Fig. 2. : Vietnam’s PISA performance (2012–2022).
**Purpose:** Historical Vietnamese PISA scores describe the setting. **Visual forms:** Line/profile plot.
**Physical source pages:** 3.
[PNG](<03_Data_Description/P058-Fig02-Vietnam%E2%80%99s%20PISA%20performance%20(2012-2022).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=121>) · Original paper and literature note are not hosted.
### P073 - Fig. 1
**Paper:** Investigating the factors influencing secondary school students’ creative thinking based on the opportunity–propensity framework: Evidence from high-performing countries and regions
**Original caption:** Fig. 1. Creative thinking performance distribution.
**Purpose:** Creative-thinking outcome distribution by sex. **Visual forms:** Density/histogram/boxplot.
**Physical source pages:** 7.
[PNG](<03_Data_Description/P073-Fig01-Creative%20thinking%20performance%20distribution.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=122>) · Original paper and literature note are not hosted.
### P078 - Fig. 2
**Paper:** Digital resource engagement, health challenges, and mathematical achievement in school children: An in-depth analysis of PISA data 2022
**Original caption:** Fig. 2. Status of using digital resources for mathematics learning.
**Purpose:** Descriptive frequencies of digital resource use. **Visual forms:** Stacked bar chart.
**Physical source pages:** 5.
[PNG](<03_Data_Description/P078-Fig02-Status%20of%20using%20digital%20resources%20for%20mathematics%20learning.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=123>) · Original paper and literature note are not hosted.
### P078 - Fig. 3
**Paper:** Digital resource engagement, health challenges, and mathematical achievement in school children: An in-depth analysis of PISA data 2022
**Original caption:** Fig. 3. Status of confidence of students in mathematics related tasks.
**Purpose:** Descriptive student confidence distribution. **Visual forms:** Stacked bar chart.
**Physical source pages:** 6.
[PNG](<03_Data_Description/P078-Fig03-Status%20of%20confidence%20of%20students%20in%20mathematics%20related%20tasks.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=124>) · Original paper and literature note are not hosted.
### P078 - Fig. 4
**Paper:** Digital resource engagement, health challenges, and mathematical achievement in school children: An in-depth analysis of PISA data 2022
**Original caption:** Fig. 4. Status of health problems.
**Purpose:** Descriptive health-problem frequency distribution. **Visual forms:** Stacked bar chart.
**Physical source pages:** 6.
[PNG](<03_Data_Description/P078-Fig04-Status%20of%20health%20problems.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=125>) · Original paper and literature note are not hosted.
### P085 - Fig. 2
**Paper:** The impact of digital resource utilization on student learning outcomes and self-efficacy across different economic contexts: A comparative analysis of PISA, 2022
**Original caption:** Fig. 2. Visual representation of country classification by GDP per capita (Source: https://www.imf.org/external/datamapper/NGDPDPC@WEO/OEMDC/ADVEC /WEOWORLD, Assisted on 2024–03–05).
**Purpose:** Defines economic-context groupings of countries in the sample. **Visual forms:** Map.
**Physical source pages:** 6.
[PNG](<03_Data_Description/P085-Fig02-Visual%20representation%20of%20country%20classification%20by%20GDP%20per%20capita%20(Source%20https%20www.imf.org%20external%20datamapper%20NGDPDPC@WEO%20OEMDC%20ADVEC%20WEOWORLD,%20Assisted%20on%202024-03-05).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=126>) · Original paper and literature note are not hosted.
### P085 - Fig. 4
**Paper:** The impact of digital resource utilization on student learning outcomes and self-efficacy across different economic contexts: A comparative analysis of PISA, 2022
**Original caption:** Fig. 4. Percentage in each category in the items related to UAS, UOS, and USL (0–5 represents the categories of items as described in description of variable section).
**Purpose:** Describes item-response category distributions. **Visual forms:** Bar chart.
**Physical source pages:** 8.
[PNG](<03_Data_Description/P085-Fig04-Percentage%20in%20each%20category%20in%20the%20items%20related%20to%20UAS,%20UOS,%20and%20USL%20(0-5%20represents%20the%20categories%20of%20items%20as%20described%20in%20description%20of%20variable%20section).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=127>) · Original paper and literature note are not hosted.
### P085 - Fig. 5
**Paper:** The impact of digital resource utilization on student learning outcomes and self-efficacy across different economic contexts: A comparative analysis of PISA, 2022
**Original caption:** Fig. 5. Percentage in each category in the items related to USF, USA, and USE (0 OR 1 to 4 OR 5 represents the categories of items as described in description of variable section).
**Purpose:** Describes support, activity and efficacy item response distributions. **Visual forms:** Bar chart.
**Physical source pages:** 8.
[PNG](<03_Data_Description/P085-Fig05-Percentage%20in%20each%20category%20in%20the%20items%20related%20to%20USF,%20USA,%20and%20USE%20(0%20OR%201%20to%204%20OR%205%20represents%20the%20categories%20of%20items%20as%20described%20in%20description%20of%20variable%20section).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=128>) · Original paper and literature note are not hosted.
### P090 - Fig. 3
**Paper:** Meta-affective progression in science: A cross-sectional study of junior high school students
**Original caption:** Fig. 3. Histogram of students' meta-affective ability distribution in (a) 7th grade and (b) 9th grade.
**Purpose:** Grade-specific distributions describe observed scale scores. **Visual forms:** Histogram, Composite.
**Physical source pages:** 5.
[PNG](<03_Data_Description/P090-Fig03-Histogram%20of%20students'%20meta-affective%20ability%20distribution%20in%20(a)%207th%20grade%20and%20(b)%209th%20grade.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=129>) · Original paper and literature note are not hosted.
### P093 - Fig. 3
**Paper:** Using machine learning algorithms to predict students’ general self-efficacy in PISA 2018
**Original caption:** Fig. 3. The histogram of self-efficacy IRT scores
**Purpose:** Describes the distribution of the modeled outcome. **Visual forms:** Density/histogram/boxplot.
**Physical source pages:** 9.
[PNG](<03_Data_Description/P093-Fig03-The%20histogram%20of%20self-efficacy%20IRT%20scores.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=130>) · Original paper and literature note are not hosted.
### P110 - Fig. 1
**Paper:** Attrition risk profiles of novice teachers based on their job demands, individual and school resources, and well-being
**Original caption:** Fig. 1. Number of participants per participating country.
**Purpose:** Participant counts by country describe sample coverage. **Visual forms:** Bar chart.
**Physical source pages:** 3.
[PNG](<03_Data_Description/P110-Fig01-Number%20of%20participants%20per%20participating%20country.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=131>) · Original paper and literature note are not hosted.
### P124 - Fig. U09
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** No original caption supplied.
**Purpose:** Distributions of pooled constructs. **Visual forms:** Density/histogram/boxplot, Composite.
**Physical source pages:** 23.
[PNG](<03_Data_Description/P124-FigU09-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=132>) · Original paper and literature note are not hosted.
### P131 - Fig. 2
**Paper:** Educational equity measurement: A comparative analysis of equity indicators
**Original caption:** Fig. 2. Kernel density estimates and histograms illustrating the distributional properties of four educational equity indicators across 37 educational systems. Clockwise from top left: (a) ESCS slope coefficient, (b) ESCS R², (c) inequality of opportunity index (IOpI), and (d) concentration index (CI). Vertical dashed lines indicate median values for each distribution. All estimates are derived from PISA 2018 data using Rubin’s combination rules for ten plausible values.
**Purpose:** Distributional properties of four equity indicators across countries. **Visual forms:** Density/histogram/boxplot, Composite.
**Physical source pages:** 8.
[PNG](<03_Data_Description/P131-Fig02-Kernel%20density%20estimates%20and%20histograms%20illustrating%20the%20distributional%20properties%20of%20four%20educational%20equity%20indicators%20across%2037%20educational%20systems.%20Clockwise%20from%20top....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=133>) · Original paper and literature note are not hosted.
### P139 - Fig. 1
**Paper:** Expecting less in hard times: How the state of the economy influences students’ educational expectations
**Original caption:** Fig. 1. GDP growth by country and year (1998 – 2018). Note: The final eight countries are presented using a different scale due to larger variance in GDP.
**Purpose:** Describes the macroeconomic covariate by country and year. **Visual forms:** Composite, Line plot.
**Physical source pages:** 4.
[PNG](<03_Data_Description/P139-Fig01-GDP%20growth%20by%20country%20and%20year%20(1998%20-%202018).%20Note%20The%20final%20eight%20countries%20are%20presented%20using%20a%20different%20scale%20due%20to%20larger%20variance%20in%20GDP.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=134>) · Original paper and literature note are not hosted.
### P144 - Fig. A.1
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure A.1: Respondents by State Notes: This figure plots a choropleth map of the geographic distribution of retained survey respondents across the contiguous United States. States are shaded by the number of retained survey respondents in each state, binned into quintiles using the empirical distribution of respondent counts across all states with at least one response. State assignments are taken from the school’s NCES state FIPS code where available and otherwise from the principal’s email domain or manual school identification. Color shading transitions from sand (lowest quintile) to maroon (highest quintile); states with no respondents are shown in light grey. The map describes raw respondent counts and is not weighted by IPW.
**Purpose:** Geographical distribution of survey respondents. **Visual forms:** Map.
**Physical source pages:** 28.
[PNG](<03_Data_Description/P144-FigA.1-Respondents%20by%20State%20Notes%20This%20figure%20plots%20a%20choropleth%20map%20of%20the%20geographic%20distribution%20of%20retained%20survey%20respondents%20across%20the%20contiguous%20United%20States.%20States%20are....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=135>) · Original paper and literature note are not hosted.
### P144 - Fig. B.4
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure B.4: Generative AI Policy Status Notes: This figure reports the IPW-weighted distribution of schools’ formal generative AI policy status as of June 2025. Principals were asked whether their school has a written policy specifically addressing generative AI use. The bars show the share of schools in each of four mutually exclusive response categories: a written policy in place at the district level, a written policy in place at the school level, a policy currently in draft, or no written policy. The figure includes all 1,254 responding principals and uses IPW weights to be nationally representative.
**Purpose:** Distribution of formal AI policy status. **Visual forms:** Bar chart.
**Physical source pages:** 34.
[PNG](<03_Data_Description/P144-FigB.4-Generative%20AI%20Policy%20Status%20Notes%20This%20figure%20reports%20the%20IPW-weighted%20distribution%20of%20schools%E2%80%99%20formal%20generative%20AI%20policy%20status%20as%20of%20June%202025.%20Principals%20were%20asked....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=136>) · Original paper and literature note are not hosted.
### P144 - Fig. B.5
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure B.5: Topics Covered by AI Policy Notes: This figure reports the IPW-weighted share of schools reporting that a district-level generative AI policy, school-level generative AI policy, or draft generative AI policy explicitly covers each topic area. Principals who reported a district-level policy, school-level policy, or draft in progress were asked, “Which topics are explicitly covered in that policy? (Select all.)” Response options are “student data privacy,” “academic honesty/plagiarism,” “special-education accommodations,” “accessibility,” “content filtering/safety,” and “other.” The sample is restricted to principals with a non-missing response to this policy-topics item. All estimates are IPW-weighted to be nationally representative of schools answering the item.
**Purpose:** Distribution of policy topic coverage. **Visual forms:** Bar chart.
**Physical source pages:** 35.
[PNG](<03_Data_Description/P144-FigB.5-Topics%20Covered%20by%20AI%20Policy%20Notes%20This%20figure%20reports%20the%20IPW-weighted%20share%20of%20schools%20reporting%20that%20a%20district-level%20generative%20AI%20policy,%20school-level%20generative%20AI....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=137>) · Original paper and literature note are not hosted.
### P144 - Fig. B.6
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure B.6: Disciplinary Actions for AI Misuse Notes: This figure reports the IPW-weighted distribution of the number of disciplinary actions for AI misuse in the most recent academic year, separately for students, teachers, and other staff, among schools reporting a district-level generative AI policy, school-level generative AI policy, or draft in progress. Principals who reported a district-level policy, school-level policy, or draft in progress were asked, “Based on existing policy at your school, what number of disciplinary actions for AI misuse occurred at your school in the most recent academic year for the following parties:” students, teachers, and other staff. Response bins are 0, 1–2, 3–5, and 6+. The sample is restricted to principals with a non-missing response to the relevant group item. All estimates are IPW-weighted to be nationally representative of schools answering the item.
**Purpose:** Distribution of disciplinary responses to AI misuse. **Visual forms:** Bar chart.
**Physical source pages:** 36.
[PNG](<03_Data_Description/P144-FigB.6-Disciplinary%20Actions%20for%20AI%20Misuse%20Notes%20This%20figure%20reports%20the%20IPW-weighted%20distribution%20of%20the%20number%20of%20disciplinary%20actions%20for%20AI%20misuse%20in%20the%20most%20recent%20academic....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=138>) · Original paper and literature note are not hosted.
### P144 - Fig. B.7
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure B.7: AI-Powered Tutoring License or Subscription Notes: This figure displays whether schools hold an active license or contract for an AI-powered tutoring platform. Principals were asked whether their school currently has a license or contract for any AI-powered tutoring platform. Response options are yes and no. All estimates are IPW-weighted to be nationally representative.
**Purpose:** Distribution of AI tutoring license uptake. **Visual forms:** Bar chart.
**Physical source pages:** 36.
[PNG](<03_Data_Description/P144-FigB.7-AI-Powered%20Tutoring%20License%20or%20Subscription%20Notes%20This%20figure%20displays%20whether%20schools%20hold%20an%20active%20license%20or%20contract%20for%20an%20AI-powered%20tutoring%20platform.%20Principals....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=139>) · Original paper and literature note are not hosted.
### P144 - Fig. B.8
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure B.8: Stance Toward Student Use of Generative AI Tools Notes: This figure reports schools’ official stance toward student use of generative AI tools. Principals were asked the current stance communicated to students regarding Generative AI use. The bars show the IPW-weighted share of schools in each of three response categories: actively encouraged for learning, including studying and homework assistance, with clear rules; discouraged but not banned; or prohibited. The figure includes all 1,254 responding principals and uses IPW weights to be nationally representative.
**Purpose:** Distribution of schools AI-use stance. **Visual forms:** Bar chart.
**Physical source pages:** 37.
[PNG](<03_Data_Description/P144-FigB.8-Stance%20Toward%20Student%20Use%20of%20Generative%20AI%20Tools%20Notes%20This%20figure%20reports%20schools%E2%80%99%20official%20stance%20toward%20student%20use%20of%20generative%20AI%20tools.%20Principals%20were%20asked%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=140>) · Original paper and literature note are not hosted.
### P144 - Fig. B.9
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure B.9: Generative AI Widening Achievement Gaps Notes: This figure reports the IPW-weighted distribution of principal agreement with the statement, “Generative AI is widening achievement gaps among students in my school.” Responses are on a five-point agreement scale: “strongly disagree,” “somewhat disagree,” “neither agree nor disagree,” “somewhat agree,” and “strongly agree.” The figure includes principals with non-missing responses to this item. All estimates are IPW-weighted to be nationally representative.
**Purpose:** Distribution of beliefs about AI widening gaps. **Visual forms:** Bar chart.
**Physical source pages:** 37.
[PNG](<03_Data_Description/P144-FigB.9-Generative%20AI%20Widening%20Achievement%20Gaps%20Notes%20This%20figure%20reports%20the%20IPW-weighted%20distribution%20of%20principal%20agreement%20with%20the%20statement,%20%E2%80%9CGenerative%20AI%20is%20widening....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=141>) · Original paper and literature note are not hosted.
### P144 - Fig. B.10
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure B.10: Generative AI Produced Improvements in Learning Notes: This figure reports the IPW-weighted distribution of principal responses to whether Generative AI tools have produced measurable improvements in student learning at their school. Principals were asked, “Do you believe Generative AI tools have produced measurable improvements in student learning at your school?” Response options are “definitely not,” “probably not,” “might or might not,” “probably yes,” and “definitely yes.” The figure includes principals with non-missing responses to this item. All estimates are IPW-weighted to be nationally representative.
**Purpose:** Distribution of perceived learning improvements. **Visual forms:** Bar chart.
**Physical source pages:** 38.
[PNG](<03_Data_Description/P144-FigB.10-Generative%20AI%20Produced%20Improvements%20in%20Learning%20Notes%20This%20figure%20reports%20the%20IPW-weighted%20distribution%20of%20principal%20responses%20to%20whether%20Generative%20AI%20tools%20have%20produced....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=142>) · Original paper and literature note are not hosted.
### P144 - Fig. B.11
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure B.11: Generative AI Services Used by Principals Notes: This figure reports the IPW-weighted share of principals frequently using each generative AI service. Principals were asked, “Which of the following services do you frequently use? Check all that apply,” with response options for OpenAI–ChatGPT, Google–Gemini, Anthropic–Claude, X–Grok, DeepSeek–R1, and Other. Free-text responses originally reported as Other were reclassified into named service categories where identifiable; residual unclassified responses are reported in the Other bar. The sample is restricted to principals who selected at least one listed or reclassified service. All estimates are IPW-weighted to be nationally representative.
**Purpose:** Distribution of AI services used by principals. **Visual forms:** Bar chart.
**Physical source pages:** 39.
[PNG](<03_Data_Description/P144-FigB.11-Generative%20AI%20Services%20Used%20by%20Principals%20Notes%20This%20figure%20reports%20the%20IPW-weighted%20share%20of%20principals%20frequently%20using%20each%20generative%20AI%20service.%20Principals%20were%20asked,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=143>) · Original paper and literature note are not hosted.
### P144 - Fig. B.12
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure B.12: Generative AI Services Used by Teachers Notes: This figure reports the IPW-weighted share of schools where principals report teachers use each generative AI service. Principals were asked, “Which of the following services do you know teachers at your school use? Check all that apply,” with response options for OpenAI–ChatGPT, Google–Gemini, Anthropic–Claude, X–Grok, DeepSeek– R1, and Other. Free-text responses originally reported as Other were reclassified into named service categories where identifiable; residual unclassified responses are reported in the Other bar. The sample is restricted to schools where principals selected at least one listed or reclassified teacher service. All estimates are IPW-weighted to be nationally representative.
**Purpose:** Distribution of AI services used by teachers. **Visual forms:** Bar chart.
**Physical source pages:** 40.
[PNG](<03_Data_Description/P144-FigB.12-Generative%20AI%20Services%20Used%20by%20Teachers%20Notes%20This%20figure%20reports%20the%20IPW-weighted%20share%20of%20schools%20where%20principals%20report%20teachers%20use%20each%20generative%20AI%20service....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=144>) · Original paper and literature note are not hosted.
### P145 - Fig. 2
**Paper:** AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting
**Original caption:** Fig. 2. AI tutor time on task: Total time students in the AI group spent interacting with the tutor. Dotted line denotes the length of the in-class active learning lesson (60 minutes).
**Purpose:** Describes actual AI-tutor time-on-task distribution. **Visual forms:** Histogram.
**Physical source pages:** 4.
[PNG](<03_Data_Description/P145-Fig02-AI%20tutor%20time%20on%20task%20Total%20time%20students%20in%20the%20AI%20group%20spent%20interacting%20with%20the%20tutor.%20Dotted%20line%20denotes%20the%20length%20of%20the%20in-class%20active%20learning%20lesson%20(60%20minutes).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=145>) · Original paper and literature note are not hosted.
### P153 - Fig. 1
**Paper:** Do university-educated families lose their edge as education expands? The withering performance and advantage of their children
**Original caption:** Figure 1. Proportion of university-educated parents in PISA between 2003 and 2022, by country. Note: the horizontal dashed line in each graph represents 50% of parents. Source: Authors’ elaboration from PISA 2003 to 2022.
**Purpose:** Historical parental-education composition across participating countries. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 22.
[PNG](<03_Data_Description/P153-Fig01-Proportion%20of%20university-educated%20parents%20in%20PISA%20between%202003%20and%202022,%20by%20country.%20Note%20the%20horizontal%20dashed%20line%20in%20each%20graph%20represents%2050%25%20of%20parents.%20Source%20Authors%E2%80%99....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=146>) · Original paper and literature note are not hosted.
### P156 - Fig. U01
**Paper:** Effects of LLM use and note-taking on reading comprehension and memory: A randomised experiment in secondary schools
**Original caption:** Appendix C. Distribution of overarching prompt types across LLM sessions for different conditions and students Specific prompt types with frequencies. Each panel represents a specific combination of condition (LLM-only or LLM + Notes) and text passage (Apartheid in South Africa or Cuban Missile Crisis). Each bar shows the number of prompts within each type for an individual LLM session, with sessions sorted in descending order by the total number of prompts and ties broken by the number of prompts within each type.
**Purpose:** Describes prompt composition by condition, passage and session. **Visual forms:** Composite, Stacked bar chart.
**Physical source pages:** 20, 21.
[PNG](<03_Data_Description/P156-FigU01-Appendix%20C.%20Distribution%20of%20overarching%20prompt%20types%20across%20LLM%20sessions%20for%20different%20conditions%20and%20students%20Specific%20prompt%20types%20with%20frequencies.%20Each%20panel%20represents....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=147>) · Original paper and literature note are not hosted.
### P156 - Fig. 5
**Paper:** Effects of LLM use and note-taking on reading comprehension and memory: A randomised experiment in secondary schools
**Original caption:** Fig. 5. Distribution of student prompts across different types, showing the percentage of students who used the prompt type at least once (blue), as their most common prompt (magenta), and as their first prompt (green). Prompt types are arranged by overall frequency. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
**Purpose:** Describes observed student prompt usage. **Visual forms:** Grouped bar chart.
**Physical source pages:** 15.
[PNG](<03_Data_Description/P156-Fig05-Distribution%20of%20student%20prompts%20across%20different%20types,%20showing%20the%20percentage%20of%20students%20who%20used%20the%20prompt%20type%20at%20least%20once%20(blue),%20as%20their%20most%20common%20prompt....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=148>) · Original paper and literature note are not hosted.
### P167 - Fig. 6
**Paper:** Imputation Strategies Under Clinical Presence: Impact on Algorithmic Fairness
**Original caption:** Figure 6: Density distributions of the generated population.
**Purpose:** Shows the generated covariate distributions by group and outcome. **Visual forms:** Line/profile plot.
**Physical source pages:** 17.
[PNG](<03_Data_Description/P167-Fig06-Density%20distributions%20of%20the%20generated%20population.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=149>) · Original paper and literature note are not hosted.
### P172 - Fig. 1a
**Paper:** Is free and reduced-price lunch a valid measure of educational disadvantage?
**Original caption:** Figure 1a Distribution of IRS-reported household income-to-poverty ratio by 8th grade NSLP enrollment in one California school district for academic years 2008-09 through 2013-14 (N =12,500). Note: Sample size, distributions, and density bandwidths have been adjusted to comply with Census Bureau disclosure requirements. Source: Linked school district administrative records, for academic years 2008-09 through 2013-14, and Internal Revenue Service 1040 Tax data, from 2008 through 2013.
**Purpose:** Compares household income-to-poverty distributions across lunch-enrollment groups. **Visual forms:** Density/histogram/boxplot.
**Physical source pages:** 19.
[PNG](<03_Data_Description/P172-Fig1a-Distribution%20of%20IRS-reported%20household%20income-to-poverty%20ratio%20by%208th%20grade%20NSLP%20enrollment%20in%20one%20California%20school%20district%20for%20academic%20years%202008-09%20through%202013-14%20(N....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=150>) · Original paper and literature note are not hosted.
### P174 - Fig. 1A–1B
**Paper:** Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach
**Original caption:** FIGURE 1A.—Kernel density estimates of men’s real log wages 1973–1992 ($1979). FIGURE 1B.—Kernel density estimates of women’s real log wages 1973–1992 ($1979).
**Purpose:** Twenty annual wage distributions for each sex describe changing inequality. **Visual forms:** Density, Composite.
**Physical source pages:** 4, 5.
[PNG](<03_Data_Description/P174-Fig1A-1B-Kernel%20density%20estimates%20of%20men%E2%80%99s%20real%20log%20wages%201973-1992%20($1979).%20FIGURE%201B.-Kernel%20density%20estimates%20of%20women%E2%80%99s%20real%20log%20wages%201973-1992%20($1979).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=151>) · Original paper and literature note are not hosted.
### P178 - Fig. 2
**Paper:** Multilevel models with multivariate mixed response types
**Original caption:** Figure 2 Frequency of fruit intake. Never (0), less than once a week (1), once a week (2), 2–4 days a week (3), 5–6 days a week (4), once a day (5) and more than once a day (6)
**Purpose:** Observed fruit-intake response distribution. **Visual forms:** Bar chart.
**Physical source pages:** 18.
[PNG](<03_Data_Description/P178-Fig02-Frequency%20of%20fruit%20intake.%20Never%20(0),%20less%20than%20once%20a%20week%20(1),%20once%20a%20week%20(2),%202-4%20days%20a%20week%20(3),%205-6%20days%20a%20week%20(4),%20once%20a%20day%20(5)%20and%20more%20than%20once%20a%20day%20(6).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=152>) · Original paper and literature note are not hosted.
### P183 - Table 1t
**Paper:** PISA 2018 in England, Northern Ireland, Scotland and Wales: Is the data really representative of all four corners of the UK?
**Original caption:** Table 1. The distribution of pupils participating in PISA across different school year groups in Scotland. Notes: Author’s calculations using the PISA 2006–2018 data for Scotland. Green shading should be read horizontally, and refers to a greater proportion of pupils belonging to that school year group.
**Purpose:** Color-coded distribution of Scottish PISA participants across school-year groups and cycles. **Visual forms:** Complex table.
**Physical source pages:** 11.
[PNG](<03_Data_Description/P183-Table1t-The%20distribution%20of%20pupils%20participating%20in%20PISA%20across%20different%20school%20year%20groups%20in%20Scotland.%20Notes%20Author%E2%80%99s%20calculations%20using%20the%20PISA%202006-2018%20data%20for%20Scotland....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=153>) · Original paper and literature note are not hosted.
### P191 - Fig. 1
**Paper:** Substitute or Complement? Generative AI Use and Student Achievement: Evidence from PISA 2025
**Original caption:** Figure 1: Average score by group of the AI-use index. Notes: Students are sorted by the unadjusted AI-use index (the first principal component of the four chatbot questions, in standard deviations) and split into groups of about 5 percent; each point is a group’s average index and score. Orange: all students; students with identical answers stay together, so this series has 18 groups, the first being the 14 percent who answer “never” to all four questions. Blue, with its 95 percent interval: excluding the 34.8 percent who give the same answer to all four (20 groups). Every group holds at least 4,800 students, so the jumps are systematic: they reflect which answer patterns fall into each group. Senate-weighted, OECD countries; Figure 3 shows the adjusted version.
**Purpose:** Describes achievement and AI-use-index patterns across groups of students. **Visual forms:** Composite, Line/profile plot.
**Physical source pages:** 5.
[PNG](<03_Data_Description/P191-Fig01-Average%20score%20by%20group%20of%20the%20AI-use%20index.%20Notes%20Students%20are%20sorted%20by%20the%20unadjusted%20AI-use%20index%20(the%20first%20principal%20component%20of%20the%20four%20chatbot%20questions,%20in%20standard....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=154>) · Original paper and literature note are not hosted.
### P195 - Fig. 3
**Paper:** Trends in Income Nonresponse Over Two Decades
**Original caption:** Fig. 3. Demographic Characteristics of Sample by Year of Interview
**Purpose:** Describes demographic composition of the repeated survey samples. **Visual forms:** Line/profile plot.
**Physical source pages:** 9.
[PNG](<03_Data_Description/P195-Fig03-Demographic%20Characteristics%20of%20Sample%20by%20Year%20of%20Interview.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=155>) · Original paper and literature note are not hosted.
### P197 - Fig. A2
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** FIGURE A2 Results of Statistics in Medicine Volume 34 review for data-generating mechanisms. Values are both frequency and %
**Purpose:** Describes simulation practices in reviewed published studies. **Visual forms:** Composite, Dot plot.
**Physical source pages:** 28.
[PNG](<03_Data_Description/P197-FigA2-Results%20of%20Statistics%20in%20Medicine%20Volume%2034%20review%20for%20data-generating%20mechanisms.%20Values%20are%20both%20frequency%20and%20%25.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=156>) · Original paper and literature note are not hosted.
### P197 - Fig. A3
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** FIGURE A3 Results of Statistics in Medicine Volume 34 review for estimands (A) and methods (B) evaluated
**Purpose:** Describes reviewed simulation targets and methods. **Visual forms:** Composite, Dot plot.
**Physical source pages:** 29.
[PNG](<03_Data_Description/P197-FigA3-Results%20of%20Statistics%20in%20Medicine%20Volume%2034%20review%20for%20estimands%20(A)%20and%20methods%20(B)%20evaluated.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=157>) · Original paper and literature note are not hosted.
### P203 - Fig. Extended Data 2
**Paper:** A Systematic Review and Meta-Analysis of Discrepancies Between Logged and Self-Reported Digital Media Use
**Original caption:** Extended Data Fig. 2 Descriptive statistics for additional post hoc moderator analyses. Note: k: number of included effect sizes. *: One study used both a built-in tool and a third-party tool
**Purpose:** Describes the evidence available for moderator analyses. **Visual forms:** Complex table.
**Physical source pages:** 2, 45.
[PNG](<03_Data_Description/P203-FigExtended%20Data%202-Extended%20Data%20Fig.%202%20Descriptive%20statistics%20for%20additional%20post%20hoc%20moderator%20analyses.%20Note%20k%20number%20of%20included%20effect%20sizes.%201%20asterisk%20One%20study%20used%20both%20a....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=757>) · Original paper and literature note are not hosted.
### P213 - Fig. 1
**Paper:** COVID-19 school closures, learning losses and intergenerational mobility
**Original caption:** Fig. 1 Secondary enrollment rates (% of gross). World Development Indicators, World Bank.
**Purpose:** Pre-pandemic secondary-enrolment trends establish the baseline data context for projected mobility losses. **Visual forms:** Line/profile plot.
**Physical source pages:** 3.
[PNG](<03_Data_Description/P213-Fig01-Secondary%20enrollment%20rates%20(%25%20of%20gross).%20World%20Development%20Indicators,%20World%20Bank.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=758>) · Original paper and literature note are not hosted.
### P213 - App Fig. A1.1
**Paper:** COVID-19 school closures, learning losses and intergenerational mobility
**Original caption:** Figure A1.1: Extent of school closures and losses in LAYS. a. School closures and LAYS by region. b. School closures over time. Source: Azevedo et al. (2022).
**Purpose:** Background empirical school-closure exposure and associated LAYS losses motivate simulation inputs. **Visual forms:** Scatter plot, Timeline, Composite.
**Physical source pages:** 2.
[PNG](<03_Data_Description/P213-App-FigA1.1-Extent%20of%20school%20closures%20and%20losses%20in%20LAYS.%20a.%20School%20closures%20and%20LAYS%20by%20region.%20b.%20School%20closures%20over%20time.%20Source%20Azevedo%20et%20al.%20(2022).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=759>) · Original paper and literature note are not hosted.
### P214 - Fig. 1
**Paper:** Developing measures of engagement in PISA
**Original caption:** Figure 1. Distribution of disengagement for selected measures. Notes: The analyses are based on data from PISA 2018. All indices are coded so that higher values indicate higher levels of disengagement; the indices were standardised to have a mean of 0 and standard deviation of 1 across the OECD student population. Rapid guessing is only available for countries/economies that administered computer-based assessment. B-S-J-Z (China) refers to the four PISA-participating provinces/municipalities of the People’s Republic of China: Beijing, Shanghai, Jiangsu and Zhejiang. Countries and economies are ranked in descending order of the level of non-response.
**Purpose:** Country-level descriptive distributions of the engagement indicators. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 21.
[PNG](<03_Data_Description/P214-Fig01-Distribution%20of%20disengagement%20for%20selected%20measures.%20Notes%20The%20analyses%20are%20based%20on%20data%20from%20PISA%202018.%20All%20indices%20are%20coded%20so%20that%20higher%20values%20indicate%20higher%20levels....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=760>) · Original paper and literature note are not hosted.
### P231 - Fig. 1
**Paper:** Inequality of opportunity in non-cognitive skills: a cross-region analysis
**Original caption:** Fig. 1 Mean score and standard deviation (SD) of non-cognitive skills in different places.
**Purpose:** Describes means and dispersion of the measured non-cognitive skills across the seven sites. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 6.
[PNG](<03_Data_Description/P231-Fig01-Mean%20score%20and%20standard%20deviation%20(SD)%20of%20non-cognitive%20skills%20in%20different%20places.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=761>) · Original paper and literature note are not hosted.
### P235 - Fig. 1
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** Figure 1: PISA worldwide percentage correct, by question
**Purpose:** Describes baseline international difficulty of the released test items. **Visual forms:** Scatter.
**Physical source pages:** 27.
[PNG](<03_Data_Description/P235-Fig01-PISA%20worldwide%20percentage%20correct,%20by%20question.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=762>) · Original paper and literature note are not hosted.
### P238 - Fig. 2
**Paper:** Profiling low-proficiency science students in the Philippines using machine learning
**Original caption:** Fig. 2 Summary of distribution of students in different science proficiency categories. Normalized Science proficiency level distribution of students (a) and distribution of students with poor and better performance in science (b). Poor performance category is for those students who belong to Level 1b and below proficiency levels, otherwise, the students are assigned to the better performance category.
**Purpose:** Science proficiency/class distribution in the analyzed sample. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 5.
[PNG](<03_Data_Description/P238-Fig02-Summary%20of%20distribution%20of%20students%20in%20different%20science%20proficiency%20categories.%20Normalized%20Science%20proficiency%20level%20distribution%20of%20students%20(a)%20and%20distribution%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=763>) · Original paper and literature note are not hosted.
### P250 - App Fig. A.1
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure A.1. Distribution of student reports about their family’s socio-economic status Source: PISA 2022 Dataset
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 53.
[PNG](<03_Data_Description/P250-AppA-FigA.1-Distribution%20of%20student%20reports%20about%20their%20family%E2%80%99s%20socio-economic%20status%20Source%20PISA%202022%20Dataset.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=764>) · Original paper and literature note are not hosted.
### P250 - App Fig. A.2
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure A.2. Distribution of 15-year-old students across categories of school disadvantage Based on school reports on the proportion of students from disadvantaged homes in the modal grade for 15-year-olds Source: PISA 2022 Dataset
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 54.
[PNG](<03_Data_Description/P250-AppA-FigA.2-Distribution%20of%2015-year-old%20students%20across%20categories%20of%20school%20disadvantage%20Based%20on%20school%20reports%20on%20the%20proportion%20of%20students%20from%20disadvantaged%20homes%20in%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=765>) · Original paper and literature note are not hosted.
## 04 Measurement and Data Quality
### P059 - Fig. 2
**Paper:** Creative learning environments and mathematics self-efficacy as predictors of mathematics achievement: Insights from PISA 2022 across three countries – Korea, Canada, and Türkiye
**Original caption:** Fig. 2. Measurement model.
**Purpose:** Measurement model of the latent constructs. **Visual forms:** Path diagram.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P059-Fig02-Measurement%20model.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=158>) · Original paper and literature note are not hosted.
### P062 - Fig. 2
**Paper:** PISA 2022 insights on intellectual curiosity, perspective-taking, and science achievement: Examining the mediating pathways
**Original caption:** Fig. 2. Multivariate outliers, Multi= 762>69… 762.00 Extremes (≥69).
**Purpose:** Mahalanobis-distance diagnostic detects multivariate outliers. **Visual forms:** Boxplot.
**Physical source pages:** 7.
[PNG](<04_Measurement_and_Data_Quality/P062-Fig02-Multivariate%20outliers,%20Multi=%20762%20greater%20than%2069...%20762.00%20Extremes%20(%E2%89%A569).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=159>) · Original paper and literature note are not hosted.
### P070 - Fig. B1
**Paper:** Creative thinking, mathematics achievement, and well-being: A multilevel analysis of Spanish students
**Original caption:** Fig. B1. Distribution of Student Mathematics Achievement (pv1math).
**Purpose:** Appendix B explicitly checks skewness and non-normality before multilevel modelling. **Visual forms:** Histogram, Density plot.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P070-FigB1-Distribution%20of%20Student%20Mathematics%20Achievement%20(pv1math).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=160>) · Original paper and literature note are not hosted.
### P070 - Fig. B2
**Paper:** Creative thinking, mathematics achievement, and well-being: A multilevel analysis of Spanish students
**Original caption:** Fig. B2. Distribution of Student Creative Thinking Performance (pv1crth_nc).
**Purpose:** Appendix B explicitly checks skewness and non-normality before multilevel modelling. **Visual forms:** Histogram, Density plot.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P070-FigB2-Distribution%20of%20Student%20Creative%20Thinking%20Performance%20(pv1crth_nc).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=161>) · Original paper and literature note are not hosted.
### P070 - Fig. B3
**Paper:** Creative thinking, mathematics achievement, and well-being: A multilevel analysis of Spanish students
**Original caption:** Fig. B3. Distribution of Student Life Satisfaction.
**Purpose:** Appendix B explicitly checks skewness and non-normality before multilevel modelling. **Visual forms:** Histogram, Density plot.
**Physical source pages:** 9.
[PNG](<04_Measurement_and_Data_Quality/P070-FigB3-Distribution%20of%20Student%20Life%20Satisfaction.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=162>) · Original paper and literature note are not hosted.
### P070 - Fig. B4
**Paper:** Creative thinking, mathematics achievement, and well-being: A multilevel analysis of Spanish students
**Original caption:** Fig. B4. Distribution of Student Stress Resistance.
**Purpose:** Appendix B explicitly checks skewness and non-normality before multilevel modelling. **Visual forms:** Histogram, Density plot.
**Physical source pages:** 9.
[PNG](<04_Measurement_and_Data_Quality/P070-FigB4-Distribution%20of%20Student%20Stress%20Resistance.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=163>) · Original paper and literature note are not hosted.
### P070 - Fig. B5
**Paper:** Creative thinking, mathematics achievement, and well-being: A multilevel analysis of Spanish students
**Original caption:** Fig. B5. Distribution of Student Creative Self-Efficacy.
**Purpose:** Appendix B explicitly checks skewness and non-normality before multilevel modelling. **Visual forms:** Histogram, Density plot.
**Physical source pages:** 10.
[PNG](<04_Measurement_and_Data_Quality/P070-FigB5-Distribution%20of%20Student%20Creative%20Self-Efficacy.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=164>) · Original paper and literature note are not hosted.
### P070 - Fig. B6
**Paper:** Creative thinking, mathematics achievement, and well-being: A multilevel analysis of Spanish students
**Original caption:** Fig. B6. Distribution of Student Socioeconomic Status (SES).
**Purpose:** Appendix B explicitly checks skewness and non-normality before multilevel modelling. **Visual forms:** Histogram, Density plot.
**Physical source pages:** 10.
[PNG](<04_Measurement_and_Data_Quality/P070-FigB6-Distribution%20of%20Student%20Socioeconomic%20Status%20(SES).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=165>) · Original paper and literature note are not hosted.
### P070 - Fig. B7
**Paper:** Creative thinking, mathematics achievement, and well-being: A multilevel analysis of Spanish students
**Original caption:** Fig. B7. Distribution of School Creative Environment.
**Purpose:** Appendix B explicitly checks skewness and non-normality before multilevel modelling. **Visual forms:** Histogram, Density plot.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P070-FigB7-Distribution%20of%20School%20Creative%20Environment.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=166>) · Original paper and literature note are not hosted.
### P070 - Fig. B8
**Paper:** Creative thinking, mathematics achievement, and well-being: A multilevel analysis of Spanish students
**Original caption:** Fig. B8. Distribution of School-Average Creative Thinking.
**Purpose:** Appendix B explicitly checks skewness and non-normality before multilevel modelling. **Visual forms:** Histogram, Density plot.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P070-FigB8-Distribution%20of%20School-Average%20Creative%20Thinking.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=167>) · Original paper and literature note are not hosted.
### P070 - Fig. B9
**Paper:** Creative thinking, mathematics achievement, and well-being: A multilevel analysis of Spanish students
**Original caption:** Fig. B9. Distribution of School-Average SES.
**Purpose:** Appendix B explicitly checks skewness and non-normality before multilevel modelling. **Visual forms:** Histogram, Density plot.
**Physical source pages:** 12.
[PNG](<04_Measurement_and_Data_Quality/P070-FigB9-Distribution%20of%20School-Average%20SES.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=168>) · Original paper and literature note are not hosted.
### P071 - Table 1
**Paper:** Effects of individual attributes, family background, and school context on students’ global competence: Insights from the OECD PISA 2018
**Original caption:** Table 1 Description of variables.
**Purpose:** Hierarchically partitioned measure definitions with corresponding questionnaire examples. **Visual forms:** Complex table.
**Physical source pages:** 4.
[PNG](<04_Measurement_and_Data_Quality/P071-Table01-Description%20of%20variables.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=169>) · Original paper and literature note are not hosted.
### P073 - Table 2
**Paper:** Investigating the factors influencing secondary school students’ creative thinking based on the opportunity–propensity framework: Evidence from high-performing countries and regions
**Original caption:** Table 2 Descriptions of the variables used in this study.
**Purpose:** Construct measurement taxonomy includes control, opportunity, creative and social-emotional partitions with reliability and sample items. **Visual forms:** Complex table.
**Physical source pages:** 5.
[PNG](<04_Measurement_and_Data_Quality/P073-Table02-Descriptions%20of%20the%20variables%20used%20in%20this%20study.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=170>) · Original paper and literature note are not hosted.
### P081 - Table 1
**Paper:** Latent classes of students' ICT engagement behaviors: Predictive role of psychological factors based on PISA 2022 data
**Original caption:** Table 1 Variables, Indicators, and Measurement Items. Table 1 (continued) Note. The response categories for selected variables were defined according to the standards of the 2022 Technical Report.
**Purpose:** Hierarchical instrument codebook maps theoretical dimensions to questionnaire items, recoding and response options. **Visual forms:** Complex table.
**Physical source pages:** 4, 5.
[PNG](<04_Measurement_and_Data_Quality/P081-Table01-Variables,%20Indicators,%20and%20Measurement%20Items.%20Table%201%20(continued)%20Note.%20The%20response%20categories%20for%20selected%20variables%20were%20defined%20according%20to%20the%20standards....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=171>) · Original paper and literature note are not hosted.
### P081 - Fig. 5
**Paper:** Latent classes of students' ICT engagement behaviors: Predictive role of psychological factors based on PISA 2022 data
**Original caption:** Fig. 5. Variable Importance in Determining Latent Class Groups Across Hong Kong (China), Japan, Korea, Singapore.
**Purpose:** Item discrimination/importance for distinguishing the latent engagement classes. **Visual forms:** Bar chart.
**Physical source pages:** 12.
[PNG](<04_Measurement_and_Data_Quality/P081-Fig05-Variable%20Importance%20in%20Determining%20Latent%20Class%20Groups%20Across%20Hong%20Kong%20(China),%20Japan,%20Korea,%20Singapore.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=172>) · Original paper and literature note are not hosted.
### P082 - App Table A1
**Paper:** C017
**Original caption:** Table A1. ICT Access Factors in PISA ICT Familiarity Questionnaires from 2000 to 2022. Note. “O” indicates that the factor was measured in the cycle, while “-” indicates that the factor was not measured.
**Purpose:** Original appendix measurement-coverage matrix across PISA cycles, not a routine descriptive-results table. **Visual forms:** Complex table, Measurement coverage matrix.
**Physical source pages:** 1.
[PNG](<04_Measurement_and_Data_Quality/P082-App-TableA1-ICT%20Access%20Factors%20in%20PISA%20ICT%20Familiarity%20Questionnaires%20from%202000%20to%202022.%20Note.%20%E2%80%9CO%E2%80%9D%20indicates%20that%20the%20factor%20was%20measured%20in%20the%20cycle,%20while%20%E2%80%9C-%E2%80%9D....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=766>) · Original paper and literature note are not hosted.
### P082 - App Table A2
**Paper:** C017
**Original caption:** Table A2. ICT Use Factors in PISA ICT Familiarity Questionnaires from 2000 to 2022. Note. “O” indicates that the factor was measured in the cycle, while “-” indicates that the factor was not measured. In the 2000 cycle, the labels “for leisure purpose (no location distinction)” and “for academic purpose (no location distinction)” were created by the author based on specific activities listed in the original questionnaire, such as playing games, word processing, spreadsheets, and graphic design. In the 2003 and 2006 cycles, original categories such as “Internet/entertainment use” and “program/software use” were reclassified by the author into “leisure” and “academic” use. Similarly, in 2012, the original categories “entertainment use” and “school-related tasks” were recoded as “leisure” and “academic” use.
**Purpose:** Hierarchical longitudinal measurement-coverage comparison, including full continuation and harmonization note. **Visual forms:** Complex table, Measurement coverage matrix.
**Physical source pages:** 2, 3.
[PNG](<04_Measurement_and_Data_Quality/P082-App-TableA2-ICT%20Use%20Factors%20in%20PISA%20ICT%20Familiarity%20Questionnaires%20from%202000%20to%202022.%20Note.%20%E2%80%9CO%E2%80%9D%20indicates%20that%20the%20factor%20was%20measured%20in%20the%20cycle,%20while%20%E2%80%9C-%E2%80%9D%20indicates....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=767>) · Original paper and literature note are not hosted.
### P082 - App Table A3
**Paper:** C017
**Original caption:** Table A3. ICT Attitude Factors in PISA ICT Familiarity Questionnaires from 2000 to 2022. Note. “O” indicates that the factor was measured in the cycle, while “-” indicates that the factor was not measured. In the 2012 cycle, the variable labeled “attitude towards computers” was originally measured using two separate constructs: “computers as a tool for school learning” and “limitations of computers as a tool for school learning.”
**Purpose:** Measurement-coverage matrix with full definition note; lower crop extended to retain all note lines. **Visual forms:** Complex table, Measurement coverage matrix.
**Physical source pages:** 4.
[PNG](<04_Measurement_and_Data_Quality/P082-App-TableA3-ICT%20Attitude%20Factors%20in%20PISA%20ICT%20Familiarity%20Questionnaires%20from%202000%20to%202022.%20Note.%20%E2%80%9CO%E2%80%9D%20indicates%20that%20the%20factor%20was%20measured%20in%20the%20cycle,%20while%20%E2%80%9C-%E2%80%9D....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=768>) · Original paper and literature note are not hosted.
### P085 - Fig. 3
**Paper:** The impact of digital resource utilization on student learning outcomes and self-efficacy across different economic contexts: A comparative analysis of PISA, 2022
**Original caption:** Fig. 3. Factor loading of measurement model of CFA.
**Purpose:** CFA factor-loadings measurement model. **Visual forms:** Path diagram.
**Physical source pages:** 7.
[PNG](<04_Measurement_and_Data_Quality/P085-Fig03-Factor%20loading%20of%20measurement%20model%20of%20CFA.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=173>) · Original paper and literature note are not hosted.
### P090 - Table 1
**Paper:** Meta-affective progression in science: A cross-sectional study of junior high school students
**Original caption:** Table 1 Sample statement of Meta-Affective Traits for Science Young Learner (MATSYL).
**Purpose:** Nested dimensions and indicators map to sample measurement statements. **Visual forms:** Complex table.
**Physical source pages:** 4.
[PNG](<04_Measurement_and_Data_Quality/P090-Table01-Sample%20statement%20of%20Meta-Affective%20Traits%20for%20Science%20Young%20Learner%20(MATSYL).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=174>) · Original paper and literature note are not hosted.
### P094 - Table 2
**Paper:** What teachers can do to enhance students’ well-being: Discussion
**Original caption:** Table 2. Well-being: Function, concept and measurement.
**Purpose:** Structured cross-study comparison of the special issue evidence, with multiple conceptual partitions. **Visual forms:** Complex table, Structured evidence comparison.
**Physical source pages:** 3.
[PNG](<04_Measurement_and_Data_Quality/P094-Table02-Well-being%20Function,%20concept%20and%20measurement.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=175>) · Original paper and literature note are not hosted.
### P097 - Table 3
**Paper:** The role of parental involvement in adolescents’ digital reading performance
**Original caption:** Table 3 Description of variables used in this study. Sources. https://www.oecd.org/pisa/data/2018database; https://www.itu.int/net4/ITU-D/idi/2017/index.html.
**Purpose:** Hierarchical variable mapping links constructs, interaction codes, outcome and three control-variable levels. **Visual forms:** Complex table.
**Physical source pages:** 6.
[PNG](<04_Measurement_and_Data_Quality/P097-Table03-Description%20of%20variables%20used%20in%20this%20study.%20Sources.%20https%20www.oecd.org%20pisa%20data%202018database;%20https%20www.itu.int%20net4%20ITU-D%20idi%202017%20index.html.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=176>) · Original paper and literature note are not hosted.
### P110 - Fig. 2
**Paper:** Attrition risk profiles of novice teachers based on their job demands, individual and school resources, and well-being
**Original caption:** Fig. 2. ω values (on the x-axis) for cluster solutions 3–14 (on the y-axis) using K-Means.
**Purpose:** Cluster-solution adequacy used to select attrition-profile measurement. **Visual forms:** Bar chart.
**Physical source pages:** 4.
[PNG](<04_Measurement_and_Data_Quality/P110-Fig02-%CF%89%20values%20(on%20the%20x-axis)%20for%20cluster%20solutions%203-14%20(on%20the%20y-axis)%20using%20K-Means.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=177>) · Original paper and literature note are not hosted.
### P116 - Fig. 2
**Paper:** School leadership and teacher stress: Unpacking the mediating role of job satisfaction and the employment status as a moderator
**Original caption:** Fig. 2. Mahalonobis distance.
**Purpose:** Mahalanobis-distance screening for multivariate outliers. **Visual forms:** Density/histogram/boxplot.
**Physical source pages:** 5.
[PNG](<04_Measurement_and_Data_Quality/P116-Fig02-Mahalonobis%20distance.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=178>) · Original paper and literature note are not hosted.
### P122 - Table 1
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Table 1. Constructs, questions and indicators from PIRLS teacher questionnaire and TALIS. Table 1 (continued).
**Purpose:** Structured operationalization maps assessment-literacy dimensions to questionnaire questions, answer scales, derived indicators, sources and years. **Visual forms:** Complex table.
**Physical source pages:** 5.
[PNG](<04_Measurement_and_Data_Quality/P122-Table01-Constructs,%20questions%20and%20indicators%20from%20PIRLS%20teacher%20questionnaire%20and%20TALIS.%20Table%201%20(continued).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=179>) · Original paper and literature note are not hosted.
### P127 - Fig. 3
**Paper:** Teacher evaluation – A resource for teaching quality or a burden? An analysis using TALIS 2018 data
**Original caption:** Fig. 3. Measurement model of the dependent variables (Model fit: RMSEA = .05; CFI = .96; TLI = .95; n = 207481; all paths p < .001).
**Purpose:** CFA measurement model of dependent constructs. **Visual forms:** Path diagram.
**Physical source pages:** 5.
[PNG](<04_Measurement_and_Data_Quality/P127-Fig03-Measurement%20model%20of%20the%20dependent%20variables%20(Model%20fit%20RMSEA%20=%20.05;%20CFI%20=%20.96;%20TLI%20=%20.95;%20n%20=%20207481;%20all%20paths%20p%20less%20than%20.001).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=180>) · Original paper and literature note are not hosted.
### P130 - Fig. X
**Paper:** A critical evaluation of alignment optimization for improving cross-national comparability in international large-scale assessments
**Original caption:** Fig. X. Measurement Invariance Level of Principal Scales in TALIS 2018 Report.
**Purpose:** Measurement-invariance levels for principal scales. **Visual forms:** Bar chart.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P130-FigX-Measurement%20Invariance%20Level%20of%20Principal%20Scales%20in%20TALIS%202018%20Report.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=181>) · Original paper and literature note are not hosted.
### P130 - Fig. Y
**Paper:** A critical evaluation of alignment optimization for improving cross-national comparability in international large-scale assessments
**Original caption:** Fig. Y. Measurement Invariance Level of Teacher Scales in TALIS 2018 Report.
**Purpose:** Measurement-invariance levels for teacher scales. **Visual forms:** Bar chart.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P130-FigY-Fig.%20Y.%20Measurement%20Invariance%20Level%20of%20Teacher%20Scales%20in%20TALIS%202018%20Report.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=182>) · Original paper and literature note are not hosted.
### P132 - Fig. 1
**Paper:** Evaluating test-taking motivation based on process data and its influence on academic performance
**Original caption:** Fig. 1. Elbow plot (a) and silhouette coefficient plot (b).
**Purpose:** Cluster-number diagnostics inform behavioral-effort measurement. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 6.
[PNG](<04_Measurement_and_Data_Quality/P132-Fig01-Elbow%20plot%20(a)%20and%20silhouette%20coefficient%20plot%20(b).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=183>) · Original paper and literature note are not hosted.
### P133 - Fig. 2
**Paper:** How justified are the criticisms of bias against the OECD’s global competence framework?
**Original caption:** Fig. 2. Path diagrams of measurement model.
**Purpose:** Measurement model used for invariance and bias analyses. **Visual forms:** Path diagram.
**Physical source pages:** 5.
[PNG](<04_Measurement_and_Data_Quality/P133-Fig02-Path%20diagrams%20of%20measurement%20model.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=184>) · Original paper and literature note are not hosted.
### P134 - Fig. 5
**Paper:** Separating prior knowledge from acquired knowledge: An individual differences analysis of PISA - learning in the digital world
**Original caption:** Fig. 5. Overview of Item Infit for the Pretest (left) and the BC (right) Dimension.
**Purpose:** Item-infit distributions diagnose measurement model fit. **Visual forms:** Histogram, Composite.
**Physical source pages:** 7.
[PNG](<04_Measurement_and_Data_Quality/P134-Fig05-Overview%20of%20Item%20Infit%20for%20the%20Pretest%20(left)%20and%20the%20BC%20(right)%20Dimension.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=185>) · Original paper and literature note are not hosted.
### P134 - Fig. 6
**Paper:** Separating prior knowledge from acquired knowledge: An individual differences analysis of PISA - learning in the digital world
**Original caption:** Fig. 6. Wright Maps showing Ability Coverage for the Pretest (left) and the BC (right) Dimension.
**Purpose:** Ability–item targeting checks coverage of the pretest and challenge scales. **Visual forms:** Composite, Wright map.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P134-Fig06-Wright%20Maps%20showing%20Ability%20Coverage%20for%20the%20Pretest%20(left)%20and%20the%20BC%20(right)%20Dimension.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=186>) · Original paper and literature note are not hosted.
### P136 - Fig. 1
**Paper:** The essence of creative thinking in PISA 2022: An item-level machine learning approach to construct validity at the facet level
**Original caption:** Fig. 1. Most important predictors. Note. GDI = generate diverse ideas, GCI = generate creative ideas, EII = elaborate and improve ideas; A full description of the most important predictor variables displayed here is presented in Appendix A.
**Purpose:** Predictor importance diagnoses construct validity of three creativity facets. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 9.
[PNG](<04_Measurement_and_Data_Quality/P136-Fig01-Most%20important%20predictors.%20Note.%20GDI%20=%20generate%20diverse%20ideas,%20GCI%20=%20generate%20creative%20ideas,%20EII%20=%20elaborate%20and%20improve%20ideas;%20A%20full%20description%20of%20the%20most....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=187>) · Original paper and literature note are not hosted.
### P136 - Fig. 2
**Paper:** The essence of creative thinking in PISA 2022: An item-level machine learning approach to construct validity at the facet level
**Original caption:** Fig. 2. Estimated relationship between domains and scored categories. Note. GDI = generate diverse ideas, GCI = generate creative ideas, EII = elaborate and improve ideas; ScPS = science problem solving, SoPS social problem solving, VE = verbal expression, WE = written expression.
**Purpose:** Scoring-category differences by task domain diagnose construct coverage. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 10.
[PNG](<04_Measurement_and_Data_Quality/P136-Fig02-Estimated%20relationship%20between%20domains%20and%20scored%20categories.%20Note.%20GDI%20=%20generate%20diverse%20ideas,%20GCI%20=%20generate%20creative%20ideas,%20EII%20=%20elaborate%20and%20improve....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=188>) · Original paper and literature note are not hosted.
### P136 - Fig. 3
**Paper:** The essence of creative thinking in PISA 2022: An item-level machine learning approach to construct validity at the facet level
**Original caption:** Fig. 3. Estimated relationship between abilities and scored categories. Note. GDI = generate diverse ideas, GCI = generate creative ideas, EII = elaborate and improve ideas.
**Purpose:** Relations of academic abilities to scoring categories diagnose construct validity. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P136-Fig03-Estimated%20relationship%20between%20abilities%20and%20scored%20categories.%20Note.%20GDI%20=%20generate%20diverse%20ideas,%20GCI%20=%20generate%20creative%20ideas,%20EII%20=%20elaborate%20and%20improve%20ideas.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=189>) · Original paper and literature note are not hosted.
### P136 - Fig. 4
**Paper:** The essence of creative thinking in PISA 2022: An item-level machine learning approach to construct validity at the facet level
**Original caption:** Fig. 4. Estimated relationship between abilities and expected scores. Note. GDI = generate diverse ideas, GCI = generate creative ideas, EII = elaborate and improve ideas.
**Purpose:** Expected creativity scores vary with measured academic abilities. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 12.
[PNG](<04_Measurement_and_Data_Quality/P136-Fig04-Estimated%20relationship%20between%20abilities%20and%20expected%20scores.%20Note.%20GDI%20=%20generate%20diverse%20ideas,%20GCI%20=%20generate%20creative%20ideas,%20EII%20=%20elaborate%20and%20improve%20ideas.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=190>) · Original paper and literature note are not hosted.
### P137 - Fig. 1
**Paper:** The role of item format in the PISA 2018 mathematics literacy assessment: A cross-country study
**Original caption:** Fig. 1. Student, Item, and Residual Variance Components (VC) for the PISA 2018 Mathematical Literacy Assessment. Note. The countries are arranged in descending order of their item variance component. The VCs correspond to the proportion of the total response variation within each country attributed to students or items. Countries that administered the alternative M6B cluster are denoted with an asterisk (*).
**Purpose:** Response variance decomposition diagnoses the contribution of items versus students. **Visual forms:** Bar chart.
**Physical source pages:** 5.
[PNG](<04_Measurement_and_Data_Quality/P137-Fig01-Student,%20Item,%20and%20Residual%20Variance%20Components%20(VC)%20for%20the%20PISA%202018%20Mathematical%20Literacy%20Assessment.%20Note.%20The%20countries%20are%20arranged%20in%20descending%20order%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=191>) · Original paper and literature note are not hosted.
### P137 - Fig. 2
**Paper:** The role of item format in the PISA 2018 mathematics literacy assessment: A cross-country study
**Original caption:** Fig. 2. Average Differences in Item Difficulty Between Items Presented in Multiple-Choice (MC; reference category) versus Constructed Response (CR) Formats. Note. The regression weights b1 from the explanatory cross-classified mixed effects models (M 1) are provided on the logit scale. The b1 regression weight reflects the expected differences in item difficulty between the constructed response (CR) item format and the multiple-choice (MC; reference category) item format. Positive and negative b1 values indicate, respectively, higher and lower odds of an average student correctly responding to CR items versus MC items. The results of the multivariate random-effects meta-analysis are displayed at the bottom, with the black diamond representing the 95 % confidence interval of the pooled average b1, while the surrounding bars define the prediction interval, reflecting the dispersion around this pooled average. The R2(̂βi) indicates the relative reduction in item variance attributable to item format. Countries that administered the alternative M6B cluster are denoted with an asterisk (*).
**Purpose:** Item-format differences in measurement difficulty across countries. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 6.
[PNG](<04_Measurement_and_Data_Quality/P137-Fig02-Average%20Differences%20in%20Item%20Difficulty%20Between%20Items%20Presented%20in%20Multiple-Choice%20(MC;%20reference%20category)%20versus%20Constructed%20Response%20(CR)%20Formats.%20Note.%20The....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=192>) · Original paper and literature note are not hosted.
### P137 - Fig. 3
**Paper:** The role of item format in the PISA 2018 mathematics literacy assessment: A cross-country study
**Original caption:** Fig. 3. Across-Countries Average Distribution of Item Difficulties in the PISA 2018 Mathematics Literacy Assessment for Multiple-Choice (MC) and Constructed Response (CR) Item Formats. Note. For each country, the item difficulties for each item were computed based on the parameter estimates from the M 1 model (Equation (2)), using b1 as the average expected difference in item difficulty between MC and CR formats and εi as the residual item difficulty, not accounted for by item format. For MC format, difficulties are given by b1 × Format(MC = 0) + εi, and for CR items by b1 × Format(CR = 1) + εi. The resulting individual items’ difficulties were then averaged across countries, and the item counts for each format per difficulty level were obtained. The difficulties on the horizontal axis are displayed on a logit scale, with CR item difficulties plotted above the horizontal axis and MC item difficulties below, facilitating direct comparison across the difficulty spectrum. Negative and positive values of the horizontal axis indicate higher and lower difficulty of items, respectively.
**Purpose:** Compares difficulty coverage of two item formats. **Visual forms:** Histogram.
**Physical source pages:** 7.
[PNG](<04_Measurement_and_Data_Quality/P137-Fig03-Across-Countries%20Average%20Distribution%20of%20Item%20Difficulties%20in%20the%20PISA%202018%20Mathematics%20Literacy%20Assessment%20for%20Multiple-Choice%20(MC)%20and%20Constructed%20Response%20(CR)....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=193>) · Original paper and literature note are not hosted.
### P137 - Fig. D1
**Paper:** The role of item format in the PISA 2018 mathematics literacy assessment: A cross-country study
**Original caption:** Fig. D1. Distributions of Item Difficulties for Multiple-Choice (MC) and Constructed Response (CR) Formats. Note. For each country, the item difficulties for each item were computed based on the parameter estimates from the M 1 model (Equation (2)), using b1 as the average expected difference in difficulty between MC and CR formats and εi as item-specific deviations, not accounted for by item format. For MC format, difficulties are given by b1× Format(MC = 0) + εi, and for CR items by b1× Format(CR = 1) + εi. From the resulting difficulties, item counts for each format per difficulty level were obtained. The difficulties on the horizontal axis are displayed on a logit scale, with CR item difficulties plotted above the horizontal axis and MC item difficulties below, facilitating direct comparison across the difficulty spectrum. Negative and positive values of the horizontal axis indicate higher and lower difficulty of items, respectively. Countries are listed in ascending order based on the gender-related difference in item format. An asterisk (*) marks countries that administered the alternative M6B cluster.
**Purpose:** Country-wise item-format difficulty coverage checks measurement comparability. **Visual forms:** Histogram, Composite.
**Physical source pages:** 13.
[PNG](<04_Measurement_and_Data_Quality/P137-FigD1-Distributions%20of%20Item%20Difficulties%20for%20Multiple-Choice%20(MC)%20and%20Constructed%20Response%20(CR)%20Formats.%20Note.%20For%20each%20country,%20the%20item%20difficulties%20for%20each%20item....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=194>) · Original paper and literature note are not hosted.
### P144 - Table A.1
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Table A.1: Survey Modules Notes: This table summarizes the eight modules of the Principals AI Survey. Each row lists the module number, title, and a brief description of the topics covered. Modules 0 and 7 collect school and principal background; Modules 1–6 cover the substantive content used throughout the analysis: decision authority and technology budget, AI policy and governance, teacher professional development, AI-assisted tutoring and instruction, student adoption and guidance, and barriers, enablers, and outlook. The analysis sample used throughout the appendix contains 1,254 retained respondents.
**Purpose:** Structured survey module-to-content measurement design. **Visual forms:** Complex table.
**Physical source pages:** 25.
[PNG](<04_Measurement_and_Data_Quality/P144-TableA.1-Survey%20Modules%20Notes%20This%20table%20summarizes%20the%20eight%20modules%20of%20the%20Principals%20AI%20Survey.%20Each%20row%20lists%20the%20module%20number,%20title,%20and%20a%20brief%20description%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=195>) · Original paper and literature note are not hosted.
### P157 - Fig. 3
**Paper:** Estimating Social and Ethnic Inequality in School Surveys: Biases from Child Misreporting and Parent Nonresponse
**Original caption:** Figure 3. Sweden. Predicted mean z-standardized test scores with cluster robust 95 per cent confidence intervals from linear regression on parental education and immigrant background (each entered as three dummy variables), controlling for sex, age, and stratum. Graphs in left panels display estimated coefficients for parental education (holding immigrant background constant at ‘none’), those in right panels the estimated immigrant–non-immigrant gap (holding parental education constant at ‘primary’). Graphs in top panels use only cases with information from all three sources (complete cases), those in bottom panels use all available cases for each source (full sample). Top panel N = 2,483; bottom panel N = 4,413 (pupils), 2,607 (parents), 4,635 (register).
**Purpose:** Compares child/parent/register measurements and complete/full samples to diagnose reporting and nonresponse biases. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P157-Fig03-Sweden.%20Predicted%20mean%20z-standardized%20test%20scores%20with%20cluster%20robust%2095%20per%20cent%20confidence%20intervals%20from%20linear%20regression%20on%20parental%20education%20and%20immigrant....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=196>) · Original paper and literature note are not hosted.
### P157 - Fig. 4
**Paper:** Estimating Social and Ethnic Inequality in School Surveys: Biases from Child Misreporting and Parent Nonresponse
**Original caption:** Figure 4. England, Germany, the Netherlands, and Sweden. Estimated gradients (with cluster robust 95 per cent confidence intervals): standardized cognitive test score regressed on parental education, immigrant background, and covariates (sex, age, and stratum). Parental education as reported by pupils, parents, and obtained from registers. SES estimates represent 5 years of education. N per country = 3,796–4,475 (full sample, pupil reports), 1,463–3,469 (full sample, parent reports), and 1,409–3,387 (complete cases).
**Purpose:** Cross-country reporting-source and sample-selection quality diagnostics. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 9.
[PNG](<04_Measurement_and_Data_Quality/P157-Fig04-England,%20Germany,%20the%20Netherlands,%20and%20Sweden.%20Estimated%20gradients%20(with%20cluster%20robust%2095%20per%20cent%20confidence%20intervals)%20standardized%20cognitive%20test%20score....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=197>) · Original paper and literature note are not hosted.
### P157 - Fig. 5
**Paper:** Estimating Social and Ethnic Inequality in School Surveys: Biases from Child Misreporting and Parent Nonresponse
**Original caption:** Figure 5. England, Germany, the Netherlands, and Sweden. Estimated gradients (with cluster robust 95 per cent confidence intervals): standardized cognitive test score regressed on parental occupation, immigrant background, and covariates (sex, age, and stratum). Parental occupation as reported by pupils and parents. SES estimates represent a move of 40 steps along the ISEI-08 scale. N per country = 3,650–4,432 (full sample, pupil reports), 1,311–2,980 (full sample, parent reports), and 1,247–2,905 (complete cases).
**Purpose:** Compares parental-occupation reporting sources and sample restrictions for measurement/nonresponse bias. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 10.
[PNG](<04_Measurement_and_Data_Quality/P157-Fig05-England,%20Germany,%20the%20Netherlands,%20and%20Sweden.%20Estimated%20gradients%20(with%20cluster%20robust%2095%20per%20cent%20confidence%20intervals)%20standardized%20cognitive%20test%20score....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=198>) · Original paper and literature note are not hosted.
### P158 - Table 1
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Table 1 Examples of the used classification for systems
**Purpose:** Operational classification by input and output freedom. **Visual forms:** Complex table, Classification matrix.
**Physical source pages:** 5.
[PNG](<04_Measurement_and_Data_Quality/P158-Table01-Examples%20of%20the%20used%20classification%20for%20systems.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=199>) · Original paper and literature note are not hosted.
### P158 - Table 2
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Table 2 Overview of variables and categories
**Purpose:** Structured multi-part review coding schema. **Visual forms:** Complex table.
**Physical source pages:** 16.
[PNG](<04_Measurement_and_Data_Quality/P158-Table02-Overview%20of%20variables%20and%20categories.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=200>) · Original paper and literature note are not hosted.
### P160 - Table 1
**Paper:** Explaining the learning decline in PISA 2025 by socioeconomic quartiles through methodological changes
**Original caption:** Table 1. Household possession items by cycle, with the 16-item common core (2022–2025) and 13-item four-cycle core (2015–2025) in the final two columns. An asterisk marks desktop and laptop computers, combined into one item in the four-cycle core. Country-specific count items ST251Q08/Q09 are excluded.
**Purpose:** Multi-section item matrix documents questionnaire changes and common measurement cores. **Visual forms:** Complex table, Item-presence matrix.
**Physical source pages:** 7.
[PNG](<04_Measurement_and_Data_Quality/P160-Table01-Household%20possession%20items%20by%20cycle,%20with%20the%2016-item%20common%20core%20(2022-2025)%20and%2013-item%20four-cycle%20core%20(2015-2025)%20in%20the%20final%20two%20columns.%20An%20asterisk....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=201>) · Original paper and literature note are not hosted.
### P160 - Fig. 1
**Paper:** Explaining the learning decline in PISA 2025 by socioeconomic quartiles through methodological changes
**Original caption:** Figure 1. Average change in mathematics scores across 36 OECD systems (2022–2025), by quartile of published and harmonized indices. Color identifies the version: blue denotes the published index (OECD) and orange the harmonized index (this paper). Solid lines show ESCS and dashed lines HOMEPOS. The harmonized HOMEPOS uses a 2PL/GPCM model calibrated once on both cycles with the 16 common items, whereas the published index changed items between cycles; the harmonized ESCS combines the three harmonized components with the OECD’s rule for the composite. Grey lines show parental occupation (HISEI, solid) and parental education (PARED, dotted) on their harmonized metric, which has no published counterpart. Quartiles hold approximately 25% of the student weight each and are computed within each system and cycle; values are simple averages of 36 OECD systems. Vertical error bars show 95% confidence intervals for each quartile’s change in mean score. Standard errors combine sampling variance from the 80 BRR replicates and imputation variance across the ten plausible values, plus the link error specified in Methods.
**Purpose:** Primary measurement-comparability audit of socioeconomic grouping and achievement trends. **Visual forms:** Line plot, Point and interval plot.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P160-Fig01-Average%20change%20in%20mathematics%20scores%20across%2036%20OECD%20systems%20(2022-2025),%20by%20quartile%20of%20published%20and%20harmonized%20indices.%20Color%20identifies%20the%20version%20blue....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=202>) · Original paper and literature note are not hosted.
### P160 - Fig. 2
**Paper:** Explaining the learning decline in PISA 2025 by socioeconomic quartiles through methodological changes
**Original caption:** Figure 2. Change 2022–2025 in mean mathematics score by decile of ESCS: published versus harmonized index. Orange lines show harmonized ESCS (this paper), which rescales the household possessions component on the 16 items present in both cycles with a single joint calibration and follows the OECD’s rule for the composite; blue lines show ESCS as published by the OECD, which does not use the same questions in 2022 and in 2025, so that its change between cycles mixes a change in students with a change of instrument. Deciles are computed within each system and cycle (D1, lowest ESCS; D10, highest); values are simple averages of 36 OECD systems. Vertical error bars show 95% confidence intervals for each decile’s change in mean score. Standard errors combine sampling variance from the 80 BRR replicates and imputation variance across the ten plausible values, plus the link error specified in Methods.
**Purpose:** Primary measurement-comparability audit of socioeconomic grouping and achievement trends. **Visual forms:** Line plot, Point and interval plot.
**Physical source pages:** 12.
[PNG](<04_Measurement_and_Data_Quality/P160-Fig02-Change%202022-2025%20in%20mean%20mathematics%20score%20by%20decile%20of%20ESCS%20published%20versus%20harmonized%20index.%20Orange%20lines%20show%20harmonized%20ESCS%20(this%20paper),%20which%20rescales%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=203>) · Original paper and literature note are not hosted.
### P160 - Fig. 3
**Paper:** Explaining the learning decline in PISA 2025 by socioeconomic quartiles through methodological changes
**Original caption:** Figure 3. Level of the socioeconomic gap in mathematics, 2015–2025: published versus harmonized indices. Difference in mean mathematics score between the top and bottom quartiles (Q4 − Q1) of ESCS (solid lines) and of HOMEPOS (dashed lines). Orange denotes the harmonized four-cycle index (this paper), which uses the 13 possessions items whose concept exists in every cycle with one calibration for the four cycles, parental education mapped to a common scale, and the OECD’s rule for the composite; blue denotes the index as published by the OECD. Quartiles hold approximately 25% of the student weight each and are computed within each system and cycle; values are simple averages of the 34 OECD systems present in all four cycles. Vertical error bars show 95% confidence intervals for each estimated Q4 − Q1 gap. Standard errors combine sampling variance from the 80 BRR replicates and imputation variance across the ten plausible values.
**Purpose:** Primary measurement-comparability audit of socioeconomic grouping and achievement trends. **Visual forms:** Line plot, Point and interval plot.
**Physical source pages:** 16.
[PNG](<04_Measurement_and_Data_Quality/P160-Fig03-Level%20of%20the%20socioeconomic%20gap%20in%20mathematics,%202015-2025%20published%20versus%20harmonized%20indices.%20Difference%20in%20mean%20mathematics%20score%20between%20the%20top%20and%20bottom....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=204>) · Original paper and literature note are not hosted.
### P160 - Fig. 4
**Paper:** Explaining the learning decline in PISA 2025 by socioeconomic quartiles through methodological changes
**Original caption:** Figure 4. Change 2022–2025 in mean mathematics score by ESCS quartile, by system. Each panel is one education system identified by its three-letter code; panels are ordered by the change in the Q4 − Q1 gap under the published ESCS, from the largest narrowing (top left) to the largest widening (bottom right). Orange lines show harmonized ESCS (this paper), which rescales the household possessions component on the 16 items present in both cycles with a single joint calibration and follows the OECD’s rule for the composite; blue lines show ESCS as published by the OECD. Points are the change in the mean score of each equal-weight quartile, computed within system and cycle; the axes are shared across panels. Standard errors of the change in the gap by system are given in Table A.4.
**Purpose:** Primary measurement-comparability audit of socioeconomic grouping and achievement trends. **Visual forms:** Composite, Line plot.
**Physical source pages:** 18.
[PNG](<04_Measurement_and_Data_Quality/P160-Fig04-Change%202022-2025%20in%20mean%20mathematics%20score%20by%20ESCS%20quartile,%20by%20system.%20Each%20panel%20is%20one%20education%20system%20identified%20by%20its%20three-letter%20code;%20panels%20are%20ordered....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=205>) · Original paper and literature note are not hosted.
### P161 - Fig. 3
**Paper:** Extramural ICT factors impact adolescents’ academic performance and well-being differently: Types of self-regulated learners also matter
**Original caption:** Fig. 3 Elbow plots of information criteria of one to six profiles. Notes. AIC = Akaike’s information criterion; BIC = Bayesian information criterion; SABIC = sample-adjusted BIC
**Purpose:** Diagnostic plots inform the learner-profile measurement solution. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 19.
[PNG](<04_Measurement_and_Data_Quality/P161-Fig03-Elbow%20plots%20of%20information%20criteria%20of%20one%20to%20six%20profiles.%20Notes.%20AIC%20=%20Akaike%E2%80%99s%20information%20criterion;%20BIC%20=%20Bayesian%20information%20criterion;%20SABIC%20=....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=206>) · Original paper and literature note are not hosted.
### P162 - Fig. 1
**Paper:** Gender differences in item nonresponse in the PISA 2018 student questionnaire
**Original caption:** Fig. 1 Average gender differences in item nonresponse propensity (female students as the reference group). Note. Countries’ parameter estimates β1 (see Eq. 1) are presented on the logit scale and correspond to the cross-classified mixed effects models’ estimates of the expected differences in item nonresponse propensity on an item of an average scale of the PISA 2018 student background questionnaire for male students, as compared to female students. Positive and negative parameter estimates correspond to greater odds of item nonresponse on an item of an average scale on the PISA 2018 student questionnaire for male and female students, respectively. At the grey dashed line at 0, neither gender has greater odds of item nonresponse. The random effects meta-analytical model results are presented at the bottom of the graph. The black diamond shape corresponds to the 95% confidence interval of β1 , the pooled average across countries, and the bars around it map the corresponding prediction interval as an indication of dispersion around this average.
**Purpose:** Analyzes questionnaire nonresponse and scale-specific response-quality disparities. **Visual forms:** Composite, Forest/coefficient plot.
**Physical source pages:** 12.
[PNG](<04_Measurement_and_Data_Quality/P162-Fig01-Average%20gender%20differences%20in%20item%20nonresponse%20propensity%20(female%20students%20as%20the%20reference%20group).%20Note.%20Countries%E2%80%99%20parameter%20estimates%20%CE%B21%20(see%20Eq.%201)%20are....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=207>) · Original paper and literature note are not hosted.
### P162 - Fig. 2
**Paper:** Gender differences in item nonresponse in the PISA 2018 student questionnaire
**Original caption:** Fig. 2 Scale-specific incremental gender differences in item nonresponse propensity as a function of the scale response format and the scale content (female students as the reference group). Note. The results are reported for five scale formats and six content areas. The estimates δ ∗ s1 are presented on the logit scale and correspond to the results of the final meta-analyses that combined the estimates for each level of both covariates (i.e., format and content) across 80 countries, resulting in one pooled estimate δ ∗ s1 for each level of each covariate. The diamond shapes represent the confidence intervals around the pooled estimate for each format and content. The bars around the diamond define the corresponding prediction interval to indicate the dispersion around this average. When the estimate is significantly different from zero, the diamond shape and bars are black, otherwise, grey. When comparing female and male students with the same item nonresponse propensity, positive and negative estimates correspond to greater odds of item nonresponse for male and female students, respectively. At the grey dashed line at 0, neither male nor female students have greater odds of item nonresponse.
**Purpose:** Analyzes questionnaire nonresponse and scale-specific response-quality disparities. **Visual forms:** Composite, Forest/coefficient plot.
**Physical source pages:** 13.
[PNG](<04_Measurement_and_Data_Quality/P162-Fig02-Scale-specific%20incremental%20gender%20differences%20in%20item%20nonresponse%20propensity%20as%20a%20function%20of%20the%20scale%20response%20format%20and%20the%20scale%20content%20(female%20students%20as....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=208>) · Original paper and literature note are not hosted.
### P162 - Fig. 3
**Paper:** Gender differences in item nonresponse in the PISA 2018 student questionnaire
**Original caption:** Fig. 3 Country-wise average item nonresponse rates. Note. Each country’s average item nonresponse rate is calculated by averaging individual students’ item nonresponse rates (i.e., the ratio of the number of items on the questionnaire to which the student did not provide a response to the number of theoretically valid responses possible on the questionnaire administered within a country) to the country level.
**Purpose:** Analyzes questionnaire nonresponse and scale-specific response-quality disparities. **Visual forms:** Bar chart.
**Physical source pages:** 23.
[PNG](<04_Measurement_and_Data_Quality/P162-Fig03-Country-wise%20average%20item%20nonresponse%20rates.%20Note.%20Each%20country%E2%80%99s%20average%20item%20nonresponse%20rate%20is%20calculated%20by%20averaging%20individual%20students%E2%80%99%20item%20nonresponse....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=209>) · Original paper and literature note are not hosted.
### P162 - Fig. 4
**Paper:** Gender differences in item nonresponse in the PISA 2018 student questionnaire
**Original caption:** Fig. 4 Scale-wise average nonresponse rates. Note. Each scale’s nonresponse rate is calculated by first averaging individual items’ nonresponse rates by scale in each country and then averaging across countries for each scale. The items are arranged in the sequence as they appear on the questionnaire.
**Purpose:** Analyzes questionnaire nonresponse and scale-specific response-quality disparities. **Visual forms:** Bar chart.
**Physical source pages:** 24.
[PNG](<04_Measurement_and_Data_Quality/P162-Fig04-Scale-wise%20average%20nonresponse%20rates.%20Note.%20Each%20scale%E2%80%99s%20nonresponse%20rate%20is%20calculated%20by%20first%20averaging%20individual%20items%E2%80%99%20nonresponse%20rates%20by%20scale%20in%20each....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=210>) · Original paper and literature note are not hosted.
### P162 - Fig. 5
**Paper:** Gender differences in item nonresponse in the PISA 2018 student questionnaire
**Original caption:** Fig. 5 Odds of item nonresponse propensity as a function of item position. Note. The parameter estimates are presented on the logit scale and correspond to the θ1 parameter estimates of the cross-classified mixed effects models (Eq. 1). One unit on the item position predictor corresponds to 50 items. For students of the same gender and with the same item nonresponse propensity, positive and negative parameter values correspond to higher and lower odds of item nonresponse as they progress through the questionnaire, respectively. At the grey dashed line, neither outcome (nonresponse or response) is more likely than the other (θ1 = 0).
**Purpose:** Analyzes questionnaire nonresponse and scale-specific response-quality disparities. **Visual forms:** Composite, Forest/coefficient plot.
**Physical source pages:** 25.
[PNG](<04_Measurement_and_Data_Quality/P162-Fig05-Odds%20of%20item%20nonresponse%20propensity%20as%20a%20function%20of%20item%20position.%20Note.%20The%20parameter%20estimates%20are%20presented%20on%20the%20logit%20scale%20and%20correspond%20to%20the%20%CE%B81....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=211>) · Original paper and literature note are not hosted.
### P162 - Fig. 6
**Paper:** Gender differences in item nonresponse in the PISA 2018 student questionnaire
**Original caption:** Fig. 6 Odds of item nonresponse propensity on scales ST006 and ST008. Note. The parameter estimates are presented on the logit scale and correspond to the θ2 parameter estimates of the cross-classified mixed effects models (Eq. 1). The scales ST006 and ST008 consisted of four items each and inquired about students’ parents’ qualifications (i.e., scale ST006: Does your mother have any of the following qualifications? ST006Q01TA(doctorate), ST006Q02TA(post-graduate), ST006Q03TA(vocational tertiary), ST006Q04TA(non-tertiary post-secondary)). A yes/no response had to be selected for each item. For students of the same gender and with the same item nonresponse propensity, positive and negative parameter estimates correspond to higher and lower odds of item nonresponse on an average item of the ST006 and ST008 scales, respectively, as compared to the odds of item nonresponse on the other items with a similar position on the questionnaire. At the grey dashed line, neither outcome (nonresponse or response) is more likely than the other (θ2 = 0).
**Purpose:** Analyzes questionnaire nonresponse and scale-specific response-quality disparities. **Visual forms:** Composite, Forest/coefficient plot.
**Physical source pages:** 26.
[PNG](<04_Measurement_and_Data_Quality/P162-Fig06-Odds%20of%20item%20nonresponse%20propensity%20on%20scales%20ST006%20and%20ST008.%20Note.%20The%20parameter%20estimates%20are%20presented%20on%20the%20logit%20scale%20and%20correspond%20to%20the%20%CE%B82%20parameter....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=212>) · Original paper and literature note are not hosted.
### P162 - Fig. 7
**Paper:** Gender differences in item nonresponse in the PISA 2018 student questionnaire
**Original caption:** Fig. 7 Scale-specific incremental gender differences in item nonresponse propensity (female students as the reference group). Note. The estimates δs1 are presented on the logit scale and correspond to the results of the meta-analyses that summarised the parameter estimates for the scale-specific incremental gender differences in item nonresponse propensity (δs1, Eq. 1) across countries for each individual scale, resulting in 71 pooled estimates δs1 (one for each scale s). Positive and negative estimates correspond to greater odds of item nonresponse on the corresponding PISA 2018 student background questionnaire scale for male and female students, respectively. At the grey dashed line at 0, neither gender has greater odds of item nonresponse.
**Purpose:** Analyzes questionnaire nonresponse and scale-specific response-quality disparities. **Visual forms:** Composite, Forest/coefficient plot.
**Physical source pages:** 27.
[PNG](<04_Measurement_and_Data_Quality/P162-Fig07-Scale-specific%20incremental%20gender%20differences%20in%20item%20nonresponse%20propensity%20(female%20students%20as%20the%20reference%20group).%20Note.%20The%20estimates%20%CE%B4s1%20are%20presented%20on....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=213>) · Original paper and literature note are not hosted.
### P162 - Fig. 8
**Paper:** Gender differences in item nonresponse in the PISA 2018 student questionnaire
**Original caption:** Fig. 8 Scale-specific incremental gender differences in item nonresponse propensity as a function of the scale response format (female students as the reference group). Note. The results are reported for five response formats. The estimates δ∗ s1 are presented on the logit scale and correspond to the results of the subgroup meta-analyses that summarised the parameter estimates for the scale-specific incremental gender differences in item nonresponse propensity (δs1 in Eq. 1) for each country as a function of scale format. The diamond shapes represent the confidence intervals around the estimate for each format within each country, and the bars around the diamond define the corresponding prediction interval for a randomly sampled scale from all scales of the same format. When the estimate is significantly different from zero, the diamond shape and bars are black; otherwise, grey. When comparing female and male students with the same item nonresponse propensity, positive and negative estimates correspond to greater odds of item nonresponse for male and female students, respectively. At the grey dashed line at 0, neither male nor female students have greater odds of item nonresponse.
**Purpose:** Analyzes questionnaire nonresponse and scale-specific response-quality disparities. **Visual forms:** Composite, Forest/coefficient plot.
**Physical source pages:** 32.
[PNG](<04_Measurement_and_Data_Quality/P162-Fig08-Scale-specific%20incremental%20gender%20differences%20in%20item%20nonresponse%20propensity%20as%20a%20function%20of%20the%20scale%20response%20format%20(female%20students%20as%20the%20reference%20group)....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=214>) · Original paper and literature note are not hosted.
### P162 - Fig. 9
**Paper:** Gender differences in item nonresponse in the PISA 2018 student questionnaire
**Original caption:** Fig. 9 Scale-specific incremental gender differences as a function of the scale content (female students as the reference group). Note. The results are reported for six content areas. The estimates δ∗ s1 are presented on the logit scale and correspond to the results of the subgroup meta-analyses that summarised the parameter estimates for the scale-specific incremental gender differences in item nonresponse propensity (δs1 in Eq. 1) for each country as a function of scale content. The diamond shapes represent the confidence intervals around the estimate for each format within each country, and the bars around the diamond define the corresponding prediction interval for a randomly sampled scale from all scales covering the same content. When the estimate is significantly different from zero, the diamond shape and bars are black; otherwise, grey. When comparing female and male students with the same item nonresponse propensity, positive and negative estimates correspond to greater odds of item nonresponse for male and female students, respectively. At the grey dashed line at 0, neither male nor female students have greater odds of item nonresponse.
**Purpose:** Analyzes questionnaire nonresponse and scale-specific response-quality disparities. **Visual forms:** Composite, Forest/coefficient plot.
**Physical source pages:** 33.
[PNG](<04_Measurement_and_Data_Quality/P162-Fig09-Scale-specific%20incremental%20gender%20differences%20as%20a%20function%20of%20the%20scale%20content%20(female%20students%20as%20the%20reference%20group).%20Note.%20The%20results%20are%20reported%20for%20six....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=215>) · Original paper and literature note are not hosted.
### P163 - Fig. 1
**Paper:** Household Surveys in Crisis
**Original caption:** Figure 1 Unit Nonresponse Rates of Major Household Surveys Sources: For CPS, see Appendix G of U.S. Census Bureau (Various years-a). For SIPP, see Source and Accuracy Statement of U.S. Census Bureau (Various years-b). For NHIS, see Table 1 of U.S. Department of Health and Human Services (2014). For CE Survey, see U.S. Department of Labor (various years). For GSS, see Table A.6 of Appendix A – Sampling Design and Weighting in Smith et al. (2013).
**Purpose:** Unit-nonresponse trends quantify household-survey data quality. **Visual forms:** Line/profile plot.
**Physical source pages:** 34.
[PNG](<04_Measurement_and_Data_Quality/P163-Fig01-Unit%20Nonresponse%20Rates%20of%20Major%20Household%20Surveys%20Sources%20For%20CPS,%20see%20Appendix%20G%20of%20U.S.%20Census%20Bureau%20(Various%20years-a).%20For%20SIPP,%20see%20Source%20and%20Accuracy....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=216>) · Original paper and literature note are not hosted.
### P163 - Fig. 2
**Paper:** Household Surveys in Crisis
**Original caption:** Figure 2 Item Nonresponse Rates in the Current Population Survey (CPS) for Transfer Programs, Calculated as Share of Dollars Reported in Survey that is Imputed
**Purpose:** Transfer-program item-nonresponse diagnostic in CPS. **Visual forms:** Line/profile plot.
**Physical source pages:** 35.
[PNG](<04_Measurement_and_Data_Quality/P163-Fig02-Item%20Nonresponse%20Rates%20in%20the%20Current%20Population%20Survey%20(CPS)%20for%20Transfer%20Programs,%20Calculated%20as%20Share%20of%20Dollars%20Reported%20in%20Survey%20that%20is%20Imputed.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=217>) · Original paper and literature note are not hosted.
### P163 - Fig. 3
**Paper:** Household Surveys in Crisis
**Original caption:** Figure 3 Item Nonresponse Rates in the Survey of Income and Program Participation (SIPP) by Transfer Program, Calculated as Share of Dollars Reported in Survey that is Imputed, Excluding Imputation using Previous Wave Information
**Purpose:** Transfer-program item-nonresponse diagnostic in SIPP. **Visual forms:** Line/profile plot.
**Physical source pages:** 36.
[PNG](<04_Measurement_and_Data_Quality/P163-Fig03-Item%20Nonresponse%20Rates%20in%20the%20Survey%20of%20Income%20and%20Program%20Participation%20(SIPP)%20by%20Transfer%20Program,%20Calculated%20as%20Share%20of%20Dollars%20Reported%20in%20Survey%20that%20is....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=218>) · Original paper and literature note are not hosted.
### P164 - Fig. 1
**Paper:** How robust are socio-economic achievement gradients using PISA data? A case study from Germany
**Original caption:** FIGURE 1 The distribution of highest parental education in Germany based on (i) students’ answers in the PISA data set, (ii) parents’ answers in the PISA data set and (iii) household data in the German socio-economic panel dataset. Categories: 0 = below ISCED 1; 1 = ISCED 1; 2 = ISCED 2; 3 = ISCED 3C, 3B; 4 = ISCED 3A, 4; 5 = ISCED 5B; 6 = ISCED 5A, 6. Sample size students = 3936. Sample size parents = 2832. The student sample is based on 3936 students from the German PISA 2012 dataset (with the parental education variable coded as described in Section 4.3). The parent sample is based on 2832 parental background questionnaires from the German PISA 2012 dataset (with the parental education variable coded as described in Section 4.3). The SOP data sample is displaying highest education in 19,629 households, which were selected according to the following criteria: (i) person was not interviewed as child or youth of the household; (ii) person was born between 1952 and 1982; and (iii) one household member lived in a household with at least one child at one time point. Minor recoding of the SOEP variable ‘pgisced97’ was necessary in order to align it with the highest parental education in PISA.
**Purpose:** External comparison checks parental education measurement and coding. **Visual forms:** Stacked bar chart.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P164-Fig01-The%20distribution%20of%20highest%20parental%20education%20in%20Germany%20based%20on%20(i)%20students%E2%80%99%20answers%20in%20the%20PISA%20data%20set,%20(ii)%20parents%E2%80%99%20answers%20in%20the%20PISA%20data%20set%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=219>) · Original paper and literature note are not hosted.
### P164 - Table 1
**Paper:** How robust are socio-economic achievement gradients using PISA data? A case study from Germany
**Original caption:** TABLE 1 Cross-tabulation of student and parent reports of highest parental education. Note: Categories: 0=below ISCED 1; 1=ISCED 1; 2=ISCED 2; 3=ISCED 3C, 3B; 4=ISCED 3A, 4; 5=ISCED 5B; 6=ISCED 5A, 6. Grey shading refers to the diagonal where parent and student responses agree.
**Purpose:** Shaded diagonal visually displays parent–student reporting agreement. **Visual forms:** Complex table, Agreement matrix.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P164-Table01-Cross-tabulation%20of%20student%20and%20parent%20reports%20of%20highest%20parental%20education.%20Note%20Categories%200=below%20ISCED%201;%201=ISCED%201;%202=ISCED%202;%203=ISCED%203C,%203B;%204=ISCED%203A,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=220>) · Original paper and literature note are not hosted.
### P164 - Fig. 2
**Paper:** How robust are socio-economic achievement gradients using PISA data? A case study from Germany
**Original caption:** FIGURE 2 Average PISA mathematics scores by parental education group before and after correcting for suspected procedural errors. Categories: 0=below ISCED 1; 1=ISCED 1; 2=ISCED 2; 3=ISCED 3C, 3B; 4=ISCED 3A, 4; 5=ISCED 5B; 6=ISCED 5A, 6. Group means were computed using plausible values and accounting for sample weights.
**Purpose:** Measures impact of correcting suspected procedural errors. **Visual forms:** Grouped bar chart.
**Physical source pages:** 10.
[PNG](<04_Measurement_and_Data_Quality/P164-Fig02-Average%20PISA%20mathematics%20scores%20by%20parental%20education%20group%20before%20and%20after%20correcting%20for%20suspected%20procedural%20errors.%20Categories%200=below%20ISCED%201;%201=ISCED%201;....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=221>) · Original paper and literature note are not hosted.
### P164 - Fig. 3
**Paper:** How robust are socio-economic achievement gradients using PISA data? A case study from Germany
**Original caption:** FIGURE 3 Missing highest parental education data in the student and parent background questionnaire. The student sample is 5001 with 381 items of non-response and 684 survey non-responses. The parental sample is 5001 with 53 items of non-response and 2116 survey non-responses. ‘Yes, not returned’ refers to where the student/parent questionnaire was not completed (survey non-response). ‘Yes, not answered’ is where the student/parent questionnaire was completed, but the questions about parental education were skipped (item non-response). ‘No’ refers to where parental education data is available.
**Purpose:** Diagnoses item versus survey nonresponse. **Visual forms:** Stacked bar chart.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P164-Fig03-Missing%20highest%20parental%20education%20data%20in%20the%20student%20and%20parent%20background%20questionnaire.%20The%20student%20sample%20is%205001%20with%20381%20items%20of%20non-response%20and%20684....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=222>) · Original paper and literature note are not hosted.
### P171 - Fig. 1
**Paper:** Is Canada really an education superpower? The impact of non-participation on results from PISA 2015
**Original caption:** Fig. 1 School-response rates in PISA 2015. Source: PISA (2015) technical report. Fig. 14.1.
**Purpose:** Response-rate acceptance regions diagnose survey data quality. **Visual forms:** Scatter.
**Physical source pages:** 9.
[PNG](<04_Measurement_and_Data_Quality/P171-Fig01-School-response%20rates%20in%20PISA%202015.%20Source%20PISA%20(2015)%20technical%20report.%20Fig.%2014.1.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=223>) · Original paper and literature note are not hosted.
### P171 - Table 2
**Paper:** Is Canada really an education superpower? The impact of non-participation on results from PISA 2015
**Original caption:** Table 2 Enrolments, exclusions and response rates in PISA 2015 (OECD countries). Both weighted and unweighted school response rates are provided (the former appear in brackets). Student response rate refers to weighted figures. Shading should be read vertically, with darker green shading indicating a ‘worse’ outcome on the measure (e.g. lower response rate, higher level of exclusions). The coverage rate of 15-year-olds (f) is computed as follows: 100%−[[100%−(a)] + (b) + [100%−weighted (d)×(e)]]
**Purpose:** Color-encoded response/coverage diagnostics compare country data quality. **Visual forms:** Complex table, Heatmap/matrix.
**Physical source pages:** 5.
[PNG](<04_Measurement_and_Data_Quality/P171-Table02-Enrolments,%20exclusions%20and%20response%20rates%20in%20PISA%202015%20(OECD%20countries).%20Both%20weighted%20and%20unweighted%20school%20response%20rates%20are%20provided%20(the%20former%20appear%20in....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=224>) · Original paper and literature note are not hosted.
### P171 - Fig. 2
**Paper:** Is Canada really an education superpower? The impact of non-participation on results from PISA 2015
**Original caption:** Fig. 2 School response rates over time in selected high-performing countries
**Purpose:** School nonresponse trends compare data coverage across countries. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 10.
[PNG](<04_Measurement_and_Data_Quality/P171-Fig02-School%20response%20rates%20over%20time%20in%20selected%20high-performing%20countries.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=225>) · Original paper and literature note are not hosted.
### P171 - Fig. 3
**Paper:** Is Canada really an education superpower? The impact of non-participation on results from PISA 2015
**Original caption:** Fig. 3 Student exclusion and non-response rates over time in selected high-performing countries
**Purpose:** Student exclusion/nonresponse trends diagnose participation bias. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 13.
[PNG](<04_Measurement_and_Data_Quality/P171-Fig03-Student%20exclusion%20and%20non-response%20rates%20over%20time%20in%20selected%20high-performing%20countries.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=226>) · Original paper and literature note are not hosted.
### P179 - Fig. 1
**Paper:** Non-Response Biases in Surveys of School Children: The Case of the English PISA Samples
**Original caption:** Figure 1: PISA maths score and KS4 total points score: responding pupils, 2003. Note: the sample used is responding pupils for whom auxiliary information could be linked – see Appendix A. The PISA maths points score is the average of the five plausible values estimated by the survey organizers for each individual.
**Purpose:** Assesses relationship between linked auxiliary attainment and PISA outcome for nonresponse analysis. **Visual forms:** Scatter.
**Physical source pages:** 35.
[PNG](<04_Measurement_and_Data_Quality/P179-Fig01-PISA%20maths%20score%20and%20KS4%20total%20points%20score%20responding%20pupils,%202003.%20Note%20the%20sample%20used%20is%20responding%20pupils%20for%20whom%20auxiliary%20information%20could%20be%20linked....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=227>) · Original paper and literature note are not hosted.
### P179 - Fig. 2
**Paper:** Non-Response Biases in Surveys of School Children: The Case of the English PISA Samples
**Original caption:** Figure 2: Mean and standard deviation of KS4 total point score. Note: School design weights are used for groups (ii) and (iii) and pupil design weights for groups (iv) and (v). The groups are defined in Table 3 and in the text.
**Purpose:** Compares response-stage groups to diagnose survey selection. **Visual forms:** Composite, Line plot.
**Physical source pages:** 36.
[PNG](<04_Measurement_and_Data_Quality/P179-Fig02-Mean%20and%20standard%20deviation%20of%20KS4%20total%20point%20score.%20Note%20School%20design%20weights%20are%20used%20for%20groups%20(ii)%20and%20(iii)%20and%20pupil%20design%20weights%20for%20groups%20(iv)%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=228>) · Original paper and literature note are not hosted.
### P179 - Fig. 3
**Paper:** Non-Response Biases in Surveys of School Children: The Case of the English PISA Samples
**Original caption:** Figure 3: Predicted probability of pupil response by KS4 point score, 2003. Note: The graph shows the predicted probability of response for a boy for KS4 points scores between the 5th and 95th percentiles of the sample based on the models for 2003 in Table 5.
**Purpose:** Models nonresponse propensity against auxiliary achievement. **Visual forms:** Line plot.
**Physical source pages:** 37.
[PNG](<04_Measurement_and_Data_Quality/P179-Fig03-Predicted%20probability%20of%20pupil%20response%20by%20KS4%20point%20score,%202003.%20Note%20The%20graph%20shows%20the%20predicted%20probability%20of%20response%20for%20a%20boy%20for%20KS4%20points%20scores....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=229>) · Original paper and literature note are not hosted.
### P181 - Fig. 2
**Paper:** Participation bias in the UK Biobank distorts genetic associations and downstream analyses
**Original caption:** Fig. 2 | Performance of the UKBB probability weights. a, Truncated (*) density curves of the normalized probability weights (win) for UKBB participants, ranging from 0.02 to 50.01. b, Standardized coefficients (and 95% confidence intervals) of variables predicting UKBB participation (HSE = 0; UKBB = 1) in univariate logistic regression models. Coefficients are provided for all UKBB participants and for males and females separately. c, Correlation coefficients among all auxiliary variables within the UKBB (obtained from weighted and unweighted analyses) and within the HSE. Highlighted in blue are results where the coefficients between the UKBB (rUKBB) and the reference sample (rHSE) deviated (rdiff > 0.05, where rdiff = |rHSE − rUKBB|). d, Percentage change (for categorical variables) and change in means as a function of weighting, obtained for a number of healthrelated UKBB phenotypes, including the auxiliary variables (blue) and variables not used to construct the weights. Percentage change was estimated as the difference between the weighted (pw) and unweighted proportion (p), divided by the unweighted value ((pw − p) / p × 100). Change in means was expressed as a standardized mean difference, estimated as the difference between the unweighted mean (m) and the weighted mean (mw), divided by the unweighted standard deviation (mw − m/s.d.).
**Purpose:** Assessment of probability-weight performance and sample representativeness. **Visual forms:** Composite.
**Physical source pages:** 3, 4.
[PNG](<04_Measurement_and_Data_Quality/P181-Fig02-Performance%20of%20the%20UKBB%20probability%20weights.%20a,%20Truncated%20(%201%20asterisk%20)%20density%20curves%20of%20the%20normalized%20probability%20weights%20(win)%20for%20UKBB%20participants,%20ranging....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=230>) · Original paper and literature note are not hosted.
### P182 - Fig. 2
**Paper:** Patterns of item nonresponse behaviour to survey questionnaires are systematic and associated with genetic loci
**Original caption:** Fig. 2 | Phenotypic (tetrachoric) correlation of item nonresponse among questions. Each question has been recoded as dichotomous (1 = PNA or IDK, 0 otherwise). Each row and column represent the same question. We considered only questions that allow both the ‘Prefer not to answer’ and the ‘I don’t know’ options. Upper triangle, ‘I don’t know’; lower triangle, ‘Prefer not to answer’. The scale of possible correlations ranges from −1 to 1.
**Purpose:** Heatmap describes correlations among questionnaire nonresponse items, a data-quality and measurement pattern. **Visual forms:** Heatmap, Matrix.
**Physical source pages:** 6.
[PNG](<04_Measurement_and_Data_Quality/P182-Fig02-Phenotypic%20(tetrachoric)%20correlation%20of%20item%20nonresponse%20among%20questions.%20Each%20question%20has%20been%20recoded%20as%20dichotomous%20(1%20=%20PNA%20or%20IDK,%200%20otherwise).%20Each%20row....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=231>) · Original paper and literature note are not hosted.
### P182 - Fig. 3
**Paper:** Patterns of item nonresponse behaviour to survey questionnaires are systematic and associated with genetic loci
**Original caption:** Fig. 3 | Bar graph of factor loadings of the questions in the PNA and IDK confirmatory factor analyses. For each item in the analysis: a, The fitted loadings of PNA responses on each latent factor in the CFA with the selected bifactor model for PNA. b, The fitted loadings in CFA of the bifactor model for IDK. Error bars indicate ±1 s.e. Loadings without bars were constrained to 0 in the corresponding CFA.
**Purpose:** Factor loadings assess the latent measurement structure for the two nonresponse response styles. **Visual forms:** Composite, Bar chart.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P182-Fig03-Bar%20graph%20of%20factor%20loadings%20of%20the%20questions%20in%20the%20PNA%20and%20IDK%20confirmatory%20factor%20analyses.%20For%20each%20item%20in%20the%20analysis%20a,%20The%20fitted%20loadings%20of%20PNA....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=232>) · Original paper and literature note are not hosted.
### P183 - Fig. 2
**Paper:** PISA 2018 in England, Northern Ireland, Scotland and Wales: Is the data really representative of all four corners of the UK?
**Original caption:** Figure 2. School response rates before and after replacement in PISA 2018.Notes: Horizontal axis refer to the ‘before replacement’ school response rate—the percentage of initially sampled schools that completed the PISA test. Figures on the vertical axis refer to the ‘after replacement’ response rate—the percentage of schools that completed the PISA test after substitute schools are included in the figures. The dark-blue area, where the ‘before replacement’ level is below 65%, means the technical standard has not been met and the country should be excluded for the PISA study. The light blue ‘acceptable’ area is where countries are fully compliant with the PISA school response rate technical standard. The ‘intermediate zone’ in the middle refers to where the OECD technical standard has not been fully met, with countries required to complete a school-level NRBA Notes: Horizontal axis refer to the ‘before replacement’ school response rate—the percentage of initially sampled schools that completed the PISA test. Figures on the vertical axis refer to the ‘after replacement’ response rate—the percentage of schools that completed the PISA test after substitute schools are included in the figures. The dark-blue area, where the ‘before replacement’ level is below 65%, means the technical standard has not been met and the country should be excluded for the PISA study. The light blue ‘acceptable’ area is where countries are fully compliant with the PISA school response rate technical standard. The ‘intermediate zone’ in the middle refers to where the OECD technical standard has not been fully met, with countries required to complete a school-level NRBA. 6 of 41 J. Jerrim © 2021 The Authors. Review of Education published by John Wiley & Sons Ltd on behalf of British Educational Research Association
**Purpose:** Shows school response rates against OECD thresholds before and after replacement. **Visual forms:** Composite.
**Physical source pages:** 6.
[PNG](<04_Measurement_and_Data_Quality/P183-Fig02-School%20response%20rates%20before%20and%20after%20replacement%20in%20PISA%202018.Notes%20Horizontal%20axis%20refer%20to%20the%20%E2%80%98before%20replacement%E2%80%99%20school%20response%20rate-the%20percentage%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=233>) · Original paper and literature note are not hosted.
### P183 - Fig. 4
**Paper:** PISA 2018 in England, Northern Ireland, Scotland and Wales: Is the data really representative of all four corners of the UK?
**Original caption:** Figure 4. The within-school and total exclusion rates in PISA 2018.Notes: Dashed red lines refer to the thresholds according to the OECD technical standards. Figures for Scotland and UK as a whole are illustrated by purple square markers. Data are presented for all countries participating in PISA 2018 with data available. Source: OECD (2019: Chapter 11). Notes: Dashed red lines refer to the thresholds according to the OECD technical standards. Figures for Scotland and UK as a whole are illustrated by purple square markers. Data are presented for all countries participating in PISA 2018 with data available. Source: OECD (2019: Chapter 11). Table 1. The distribution of pupils participating in PISA across different school year groups in Scotland Notes: Author’s calculations using the PISA 2006–2018 data for Scotland. Green shading should be read horizontally, and refers to a greater proportion of pupils belonging to that school year group. PISA 2018 across the UK 11 of 41 © 2021 The Authors. Review of Education published by John Wiley & Sons Ltd on behalf of British Educational Research Association
**Purpose:** Shows within-school and total exclusion rates against technical-standard thresholds. **Visual forms:** Line/profile plot.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P183-Fig04-The%20within-school%20and%20total%20exclusion%20rates%20in%20PISA%202018.Notes%20Dashed%20red%20lines%20refer%20to%20the%20thresholds%20according%20to%20the%20OECD%20technical%20standards.%20Figures%20for....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=234>) · Original paper and literature note are not hosted.
### P183 - Fig. 6
**Paper:** PISA 2018 in England, Northern Ireland, Scotland and Wales: Is the data really representative of all four corners of the UK?
**Original caption:** Figure 6. Pupil response rates in PISA 2018.Notes: Dashed red line refers to the minimum threshold set by the technical standard. Green line indicates the OECD median pupil response rate Figures for the UK as a whole and Scotland are illustrated using a blue square. Notes: Dashed red line refers to the minimum threshold set by the technical standard. Green line indicates the OECD median pupil response rate. Figures for the UK as a whole and Scotland are illustrated using a blue square. PISA 2018 across the UK 15 of 41 © 2021 The Authors. Review of Education published by John Wiley & Sons Ltd on behalf of British Educational Research Association
**Purpose:** Shows pupil response rates against the technical minimum and OECD median. **Visual forms:** Line/profile plot.
**Physical source pages:** 15.
[PNG](<04_Measurement_and_Data_Quality/P183-Fig06-Pupil%20response%20rates%20in%20PISA%202018.Notes%20Dashed%20red%20line%20refers%20to%20the%20minimum%20threshold%20set%20by%20the%20technical%20standard.%20Green%20line%20indicates%20the%20OECD%20median%20pupil....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=235>) · Original paper and literature note are not hosted.
### P190 - Fig. 1
**Paper:** Student test-taking effort in low-stakes assessments: evidence from the English version of the PISA 2015 science test
**Original caption:** Fig. 1 RT (trimmed at 99%) distribution of item S131Q02 for the students who were administered the item in the 4th cluster position. a: Histogram of RT distribution (1 s intervals); the red vertical line corresponds to the threshold estimated by means of the FMM model. b: Densities estimated in the 2-class FMM model.
**Purpose:** Illustrates the item response-time threshold and the two-class finite-mixture decomposition. **Visual forms:** Line/profile plot.
**Physical source pages:** 6.
[PNG](<04_Measurement_and_Data_Quality/P190-Fig01-RT%20(trimmed%20at%2099%25)%20distribution%20of%20item%20S131Q02%20for%20the%20students%20who%20were%20administered%20the%20item%20in%20the%204th%20cluster%20position.%20a%20Histogram%20of%20RT%20distribution%20(1%20s....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=236>) · Original paper and literature note are not hosted.
### P191 - Fig. E.1
**Paper:** Substitute or Complement? Generative AI Use and Student Achievement: Evidence from PISA 2025
**Original caption:** Figure E.1: Loadings and variance explained, weighted principal components. Notes: OECD countries, senate-weighted principal components of the item correlation matrix. Panels (a) and (b) give each item’s loading, its correlation with the component, on the first (filled) and second (hollow) components; the broad set adds the two AI-tool items from the optional questionnaire. Panel (c) gives the share of variance each component explains. The chatbot items load almost equally on the first component, which is what makes it close to the item sum and therefore exposed to the response-style problem of Section 3 of the paper.
**Purpose:** Shows component loadings and explained variance used to evaluate the AI-use measure. **Visual forms:** Composite, Scatter, Line/profile plot.
**Physical source pages:** 17.
[PNG](<04_Measurement_and_Data_Quality/P191-App-FigE.1-Loadings%20and%20variance%20explained,%20weighted%20principal%20components.%20Notes%20OECD%20countries,%20senate-weighted%20principal%20components%20of%20the%20item%20correlation%20matrix....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=237>) · Original paper and literature note are not hosted.
### P193 - Fig. 1
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Fig. 1 Tertiary attainment of fathers (PISA) and of 35–44 year-old men (EAG). Pearson’s linear correlation coefficient between the two series is 0.78 (N=41) (Source: OECD, PISA 2015 Database; Education at a Glance Database (https://stats.oecd.org/OECDStat_Metadata/ShowMetadata.ashx?Dataset=EAG_NEAC))
**Purpose:** External comparison validates the parental-education measure. **Visual forms:** Scatter.
**Physical source pages:** 10.
[PNG](<04_Measurement_and_Data_Quality/P193-Fig01-Tertiary%20attainment%20of%20fathers%20(PISA)%20and%20of%2035-44%20year-old%20men%20(EAG).%20Pearson%E2%80%99s%20linear%20correlation%20coefficient%20between%20the%20two%20series%20is%200.78%20(N=41)%20(Source....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=238>) · Original paper and literature note are not hosted.
### P193 - Table 1
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Table 1 International household possession items across survey cycles. Source: PISA 2000–2018 questionnaires. C: included, but treated as country-specific item in the scaling of HOMEPOS; CC: in 2006, all items were treated as country-specific, but item parameters for a subset of 10 items—those marked CC and the “books at home” item—were constrained to sum to 0 for all countries. In 2009, again, all items were treated as country-specific, but all “common items” were used to determine the country means on a common scale, and a linear transformation based on these means was applied to within-country scale scores to make them “comparable”. S: Included in the questionnaire, but suppressed in the database; D: included in the questionnaire and database, but not in the scaling of the HOMEPOS index; N: not included in the questionnaire. ᵃ In 2000, the source version for this question was “In your home, do you have…”; in 2003, the source version for this question was “Which of the following do you have in your home?” and the answer format was different from that of all other years (only a tick box for “yes” was provided, meaning that missing answers could not be distinguished from “no” answers”). ᵇ In 2006, “A player (and treated as country-specific in scaling). ᶜ In 2000, the source version for the second question was “How many of these do you have at your home”. All items were in singular (“Television”, etc.). ᵈ In 2000 and 2003, the English source version used “Motor car”. ᵉ In 2000 and 2003, the English source version used “bathroom”; translation notes specified that translations of “bathroom” should refer to a place that contains washing facilities such as a shower or bathtub. ᶠ Note that this was considered a new item in 2015, when the wording was changed to include the parenthesis.
**Purpose:** Cross-cycle item/coding map documents construction and comparability of the possession component. **Visual forms:** Complex table, Matrix.
**Physical source pages:** 21, 22.
[PNG](<04_Measurement_and_Data_Quality/P193-Table01-International%20household%20possession%20items%20across%20survey%20cycles.%20Source%20PISA%202000-2018%20questionnaires.%20C%20included,%20but%20treated%20as%20country-specific%20item%20in%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=239>) · Original paper and literature note are not hosted.
### P193 - Fig. 2
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Fig. 2 Percentage of respondents with missing parental education component. The figure represents the median, interquartile range (shaded area) and adjacent values for country-level missing rates in PARED. Adjacent values are defined as the most extreme missing rates within 1.5 times the interquartile range from the nearest quartile. The following countries have missing rates that exceed the upper adjacent value: 2003: Canada (10.9), Germany (11.0), Luxembourg (13.1), New Zealand (11.9); 2006: Germany (6.2), United Kingdom (8.7), Israel (7.2), Luxembourg (6.8), New Zealand (7.9); 2009: Germany (13.0), United Kingdom (8.0), New Zealand (9.3), Panama (16.7); 2012: Germany (21.5), United Kingdom (7.6), Luxembourg (6.8), New Zealand (8.9); 2015: Brazil (6.6), Germany (17.5), United Kingdom (9.3), New Zealand (8.9) (Source: PISA 2003–2015 database)
**Purpose:** Missing parental-education rates diagnose data quality across cycles. **Visual forms:** Boxplot.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P193-Fig02-Percentage%20of%20respondents%20with%20missing%20parental%20education%20component.%20The%20figure%20represents%20the%20median,%20interquartile%20range%20(shaded%20area)%20and%20adjacent%20values%20for....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=240>) · Original paper and literature note are not hosted.
### P193 - Table 2
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Table 2 “Books at home” across survey cycles. Source: PISA 2000–2018 questionnaires; PISA 2003–2015 technical reports. When scaling HOMEPOS, in PISA 2003, only two categories (up to 100, 101 and more) were used; in PISA 2006, only three categories (“0–25 books”, “26–100 books”, “More than 100 books”) were used; in PISA 2009, only four categories (“0–25 books”, “26–100 books”, “100–500”, “More than 500 books”) were used. In PISA 2012 and 2015, all categories were used; however, for the purpose of trend scaling, in PISA 2015 only four categories (“0–10 books”, “11–100 books”, “100–500”, “More than 500 books”) were used.
**Purpose:** Cross-cycle coding crosswalk explains changes in the books-at-home measure. **Visual forms:** Complex table, Matrix.
**Physical source pages:** 24.
[PNG](<04_Measurement_and_Data_Quality/P193-Table02-%E2%80%9CBooks%20at%20home%E2%80%9D%20across%20survey%20cycles.%20Source%20PISA%202000-2018%20questionnaires;%20PISA%202003-2015%20technical%20reports.%20When%20scaling%20HOMEPOS,%20in%20PISA%202003,%20only%20two....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=241>) · Original paper and literature note are not hosted.
### P193 - Fig. 3
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Fig. 3 Percentage of respondents with missing ISCO codes for parental occupation. Share of missing values for fathers’ (left) and mothers’ (right) occupation codes. The figure represents the median, interquartile range (shaded area) and adjacent values for country-level missing rates in OCOD2 (Father ISCO code) and OCOD1 (Mother ISCO code). Adjacent values are defined as the most extreme missing rates within 1.5 times the interquartile range from the nearest quartile. The following countries have missing rates that exceed the upper adjacent value: Fathers: 2000: Israel (31.1), Japan (65.0), United States (28.3); 2003: Japan (25.4); 2006: Israel (21.7), Japan (22.0), Qatar (44.8); 2009: Georgia (30.3), Panama (25.3), Himachal Pradesh (India) (32.8), Tamil Nadu (India) (26.3); 2012: Albania (31.7), Austria (100), Japan (26.8); 2015: none. Mothers: 2000: Brasil (26.7), Israel (25.5), Japan (66.5), United States (23.1); 2003: Canada (16.4), Germany (16.7), Japan (21.3), Netherlands (15.8), New Zealand (17.8); 2006: Israel (16.5), Japan (16.3), Qatar (31.4); 2009: Canada (16.8), Germany (16.7), Japan (16.2), QTN (24.8); 2012: Albania (27.4), Austria (100), Germany (22.3); 2015: Algeria (74.8), United Kingdom (28.3), Chinese Taipei (37.0), Thailand (29.1), Tunisia (66.2) (Source: PISA 2000–2015 database)
**Purpose:** Missing occupational codes diagnose parent-specific reporting problems. **Visual forms:** Boxplot, Composite.
**Physical source pages:** 18.
[PNG](<04_Measurement_and_Data_Quality/P193-Fig03-Percentage%20of%20respondents%20with%20missing%20ISCO%20codes%20for%20parental%20occupation.%20Share%20of%20missing%20values%20for%20fathers%E2%80%99%20(left)%20and%20mothers%E2%80%99%20(right)%20occupation%20codes.%20The....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=242>) · Original paper and literature note are not hosted.
### P193 - Fig. 4
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Fig. 4 Percentage of respondents with missing parental occupation component of ESCS. The figure represents the median, interquartile range (shaded area) and adjacent values for HISEI. Adjacent values are defined as the most extreme missing rates within 1.5 times the interquartile range from the nearest quartile. The following countries have missing rates that exceed the upper adjacent value: 2000: Israel (26.0), Japan (62.5), United States (14.6); 2003: New Zealand (14.1); 2006: Azerbaidjan (17.0), Israel (16.1), Jordan (20.7), Qatar (41.2); 2009: Georgia (22.4), Himachal Pradesh (India) (32.6), Panama (17.0), Qatar (19.4), Tamil Nadu (India) (23.4); 2012: Germany (19.5), Jordan (21.8), Montenegro (16.2), Qatar (18.3), United Arab Emirates (15.0). 2015: Algeria (23.4), Thailand (22.8) (Source: PISA 2000–2015 database)
**Purpose:** Missing HISEI data diagnose the occupation component of ESCS. **Visual forms:** Boxplot.
**Physical source pages:** 18.
[PNG](<04_Measurement_and_Data_Quality/P193-Fig04-Percentage%20of%20respondents%20with%20missing%20parental%20occupation%20component%20of%20ESCS.%20The%20figure%20represents%20the%20median,%20interquartile%20range%20(shaded%20area)%20and%20adjacent....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=243>) · Original paper and literature note are not hosted.
### P193 - Table 4
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Table 4 Alternative measures of the “parental education” component of socio-economic status. A comparison based on PISA 2015 data (Public use files, excluding Albania, Spanish regions’ sample, Massachusetts, North Carolina, Puerto Rico). Country-level comparisons were conducted separately on each national sample. Only the OECD average and the country with the highest and lowest difference are included in this table.
**Purpose:** Measurement alternatives compare conversion/filter definitions and student/country validity diagnostics. **Visual forms:** Complex table.
**Physical source pages:** 31, 32.
[PNG](<04_Measurement_and_Data_Quality/P193-Table04-Alternative%20measures%20of%20the%20%E2%80%9Cparental%20education%E2%80%9D%20component%20of%20socio-economic%20status.%20A%20comparison%20based%20on%20PISA%202015%20data%20(Public%20use%20files,%20excluding%20Albania,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=244>) · Original paper and literature note are not hosted.
### P193 - Fig. 5
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Fig. 5 Correlation of average HOMEPOS (PISA) and GNI per capita (World Bank). Pearson’s linear correlation coefficient between the two series is equal to 0.65 (0.80 when GNI per capita is measured on a logarithmic scale) (Source: PISA 2015 Database and https://data.worldbank.org)
**Purpose:** External income benchmark validates household-possession scaling. **Visual forms:** Scatter.
**Physical source pages:** 23.
[PNG](<04_Measurement_and_Data_Quality/P193-Fig05-Correlation%20of%20average%20HOMEPOS%20(PISA)%20and%20GNI%20per%20capita%20(World%20Bank).%20Pearson%E2%80%99s%20linear%20correlation%20coefficient%20between%20the%20two%20series%20is%20equal%20to%200.65%20(0.80%20when....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=245>) · Original paper and literature note are not hosted.
### P193 - Table 5
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Table 5 Alternative measures of the “parental occupation” component of socio-economic status. Source: PISA 2015 database. A comparison based on PISA 2015 data (Public use files, excluding Albania, Spanish regions’ sample, Massachusetts, North Carolina, Puerto Rico). The Public Use File does not contain full ISCO codes for Austria. For this reason, the “replica” derived according to the original rules from public use files is different from the original HISEI variable in the database.
**Purpose:** Occupation-index alternatives compare coding accuracy, eligibility and validity across analysis levels. **Visual forms:** Complex table.
**Physical source pages:** 32, 33, 34.
[PNG](<04_Measurement_and_Data_Quality/P193-Table05-Alternative%20measures%20of%20the%20%E2%80%9Cparental%20occupation%E2%80%9D%20component%20of%20socio-economic%20status.%20Source%20PISA%202015%20database.%20A%20comparison%20based%20on%20PISA%202015%20data%20(Public....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=246>) · Original paper and literature note are not hosted.
### P193 - Fig. 6
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Fig. 6 Students with low household possessions and international poverty rates. The smaller, shaded chart shows an enlargement of the shaded area in the larger chart. Pearson’s linear correlation coefficient between the two series is 0.85. Values for Australia, Austria, Belgium, Canada, Cyprus, the Czech Republic, Germany, Denmark, Finland, France, Ireland, Iceland, Luxembourg, Malta, the Netherlands, Norway, Slovenia, Sweden, Switzerland and the United Kingdom, all countries with a poverty headcount ratio at or below 1%, are also shown but not labelled in the charts (Source: PISA 2015 database and https://data.worldbank.org)
**Purpose:** External poverty benchmark validates low household-possession thresholds. **Visual forms:** Scatter.
**Physical source pages:** 23.
[PNG](<04_Measurement_and_Data_Quality/P193-Fig06-Students%20with%20low%20household%20possessions%20and%20international%20poverty%20rates.%20The%20smaller,%20shaded%20chart%20shows%20an%20enlargement%20of%20the%20shaded%20area%20in%20the%20larger%20chart....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=247>) · Original paper and literature note are not hosted.
### P193 - Table 6
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Table 6 Alternative measures of the “income/wealth” component of socio-economic status. A comparison based on PISA 2015 data (Public use files, excluding Albania, Spanish regions’ sample, Massachusetts, North Carolina, Puerto Rico). Alternative 1 differs from the original variable in the database due to the use of a different scaling software [R package TAM; (Robitzsch et al. 2019)] and of a different calibration sample (in particular, different weights for Belgium and the United Kingdom, and different observations for Spain). ᵃ Also one item for Puerto Rico and one item for Albania.
**Purpose:** Household-possession variants compare item-parameter schemes, supplemental inputs and external validity. **Visual forms:** Complex table.
**Physical source pages:** 34, 35.
[PNG](<04_Measurement_and_Data_Quality/P193-Table06-Alternative%20measures%20of%20the%20%E2%80%9Cincome%20wealth%E2%80%9D%20component%20of%20socio-economic%20status.%20A%20comparison%20based%20on%20PISA%202015%20data%20(Public%20use%20files,%20excluding%20Albania,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=248>) · Original paper and literature note are not hosted.
### P193 - Fig. 7
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Fig. 7 Percentage of respondents with missing household possessions component. The figure represents the median, interquartile range (shaded area) and adjacent values for country-level missing rates in the household possession index (HOMEPOS). Adjacent values are defined as the most extreme missing rates within 1.5 times the interquartile range from the nearest quartile. The following countries have missing rates that exceed the upper adjacent value: 2003: Brasil (2.11), Canada (9.78), Czech Republic (4.14), Germany (5.67), United Kingdom (3.72), Netherlands (3.16); 2006: Bulgaria (2.06), Canada (3.53), Chile (2.05), Germany (3.15), Israel (4.03), Norway (1.98), Qatar (4.47); 2009: Brasil (5.68), Germany (7.04), Panama (6.43); 2012: Albania (6.90), Germany (15.13), Iceland (2.73), Israel (2.86), Qatar (3.24); 2015: Brazil (5.22), Germany (12.64), Russia (4.07), Tunisia (4.17) (Source: PISA 2000–2015 database)
**Purpose:** Missing household-possession rates diagnose component quality. **Visual forms:** Boxplot.
**Physical source pages:** 24.
[PNG](<04_Measurement_and_Data_Quality/P193-Fig07-Percentage%20of%20respondents%20with%20missing%20household%20possessions%20component.%20The%20figure%20represents%20the%20median,%20interquartile%20range%20(shaded%20area)%20and%20adjacent%20values....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=249>) · Original paper and literature note are not hosted.
### P193 - Table 7
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Table 7 Alternative composite measures of socio-economic status. A comparison based on PISA 2015 data (Public use files, excluding Albania, Spanish regions’ sample, Massachusetts, North Carolina, Puerto Rico).
**Purpose:** Composite-index alternatives compare component definitions, imputation, weighting and measurement consequences. **Visual forms:** Complex table.
**Physical source pages:** 36.
[PNG](<04_Measurement_and_Data_Quality/P193-Table07-Alternative%20composite%20measures%20of%20socio-economic%20status.%20A%20comparison%20based%20on%20PISA%202015%20data%20(Public%20use%20files,%20excluding%20Albania,%20Spanish%20regions%E2%80%99%20sample,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=250>) · Original paper and literature note are not hosted.
### P193 - Fig. 8
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Fig. 8 The problem with PISA 2015 paper-based instruments. The data-entry subscripts next to answer boxes are intended to help coders during data entry; they contradict however the labels provided above for respondents, leading to possible confusion. In PISA 2000, 2006, 2009 and 2012, only the last answer box (“three or more”) had a data-entry subscript
**Purpose:** Questionnaire reproduction identifies coding-label conflict and measurement error. **Visual forms:** Interface.
**Physical source pages:** 26.
[PNG](<04_Measurement_and_Data_Quality/P193-Fig08-The%20problem%20with%20PISA%202015%20paper-based%20instruments.%20The%20data-entry%20subscripts%20next%20to%20answer%20boxes%20are%20intended%20to%20help%20coders%20during%20data%20entry;%20they%20contradict....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=251>) · Original paper and literature note are not hosted.
### P193 - Fig. 9
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Fig. 9 Trends in the possession of distinct household possession items, by national income. Lower income countries include countries with a gross national income below USD 28,500 (PPP). This threshold was chosen in order to have an about equal number of countries in the lower- and higher-income group. Only countries/ economies that participated in every PISA cycle are included in the analysis. The lower values (compared to a linear trend line) observed in 2003 for many items may be related to the different response format used in that year, which did not distinguish item-non-response from a “no” answer (see notes under Table 1)
**Purpose:** Item possession trends assess cross-cycle and cross-income measurement comparability. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 28.
[PNG](<04_Measurement_and_Data_Quality/P193-Fig09-Trends%20in%20the%20possession%20of%20distinct%20household%20possession%20items,%20by%20national%20income.%20Lower%20income%20countries%20include%20countries%20with%20a%20gross%20national%20income%20below....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=252>) · Original paper and literature note are not hosted.
### P193 - Fig. Box 1
**Paper:** The measure of socio-economic status in PISA: a review and some suggested improvements
**Original caption:** Box 1 Questions about mothers’ educational qualifications. Source: PISA 2000–2018 Questionnaires. ⁸ In PISA 2015, “response” was used instead of “box”.
**Purpose:** Cross-cycle questionnaire wording and response-format comparisons explain parental-education measurement changes. **Visual forms:** Interface, Composite.
**Physical source pages:** 12, 13, 14.
[PNG](<04_Measurement_and_Data_Quality/P193-FigBox%201-Box%201%20Questions%20about%20mothers%E2%80%99%20educational%20qualifications.%20Source%20PISA%202000-2018%20Questionnaires.%208%20In%20PISA%202015,%20%E2%80%9Cresponse%E2%80%9D%20was%20used%20instead%20of%20%E2%80%9Cbox%E2%80%9D.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=253>) · Original paper and literature note are not hosted.
### P195 - Fig. 1
**Paper:** Trends in Income Nonresponse Over Two Decades
**Original caption:** Fig. 1. Plot of Item Nonresponse Rates to Income Questions by Year of Interview. Note: The top line shows nonresponse to the initial open-ended income question and the bottom line is the percent of nonresponse to bracketed income questions
**Purpose:** Diagnoses income-item nonresponse over time. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<04_Measurement_and_Data_Quality/P195-Fig01-Plot%20of%20Item%20Nonresponse%20Rates%20to%20Income%20Questions%20by%20Year%20of%20Interview.%20Note%20The%20top%20line%20shows%20nonresponse%20to%20the%20initial%20open-ended%20income%20question%20and%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=254>) · Original paper and literature note are not hosted.
### P195 - Fig. 2
**Paper:** Trends in Income Nonresponse Over Two Decades
**Original caption:** Fig. 2. Percent Income Nonresponse Reduction due to Unfolding Bracket Technique
**Purpose:** Evaluates reduction in missing income answers due to unfolding brackets. **Visual forms:** Line/profile plot.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P195-Fig02-Percent%20Income%20Nonresponse%20Reduction%20due%20to%20Unfolding%20Bracket%20Technique.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=255>) · Original paper and literature note are not hosted.
### P195 - Table 3
**Paper:** Trends in Income Nonresponse Over Two Decades
**Original caption:** Table 3. Income Nonresponse Rates Reported on Other Surveys Note: *Current Population Survey; †National Immunization Survey; ‡Consumer Expenditure Quarterly Survey; a Atrostic and Kalenkoski (2002); b Moore et al. (1999); c Dixon (2005); d Olson et al. (1999); e Battaglia et al. (2002); f McGrath (2005).
**Purpose:** Compares measurement/nonresponse procedures and outcomes across survey programs. **Visual forms:** Complex table.
**Physical source pages:** 14.
[PNG](<04_Measurement_and_Data_Quality/P195-Table03-Income%20Nonresponse%20Rates%20Reported%20on%20Other%20Surveys%20Note%201%20asterisk%20Current%20Population%20Survey;%20%E2%80%A0National%20Immunization%20Survey;%20%E2%80%A1Consumer%20Expenditure%20Quarterly....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=256>) · Original paper and literature note are not hosted.
### P197 - Fig. A1
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** FIGURE A1 Reviewer agreement on key variables for Statistics in Medicine Volume 34 review. Frequency of agreement of TPM with IRW (marker W) and MJC (marker C). For the same frequency, C is nudged left and W right to avoid visual clash [Colour figure can be viewed at wileyonlinelibrary.com]
**Purpose:** Assesses inter-reviewer coding agreement. **Visual forms:** Dot plot.
**Physical source pages:** 27.
[PNG](<04_Measurement_and_Data_Quality/P197-FigA1-Reviewer%20agreement%20on%20key%20variables%20for%20Statistics%20in%20Medicine%20Volume%2034%20review.%20Frequency%20of%20agreement%20of%20TPM%20with%20IRW%20(marker%20W)%20and%20MJC%20(marker%20C).%20For%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=257>) · Original paper and literature note are not hosted.
### P198 - Fig. 1
**Paper:** Validation of survey effort measures of grit and self-control in a sample of high school students
**Original caption:** Fig 1. Distribution of item non-response rates among those leaving questions blank. https://doi.org/10.1371/journal.pone.0235396.g001
**Purpose:** Describes the nonresponse-based effort measure under validation. **Visual forms:** Histogram.
**Physical source pages:** 5.
[PNG](<04_Measurement_and_Data_Quality/P198-Fig01-Distribution%20of%20item%20non-response%20rates%20among%20those%20leaving%20questions%20blank.%20https%20doi.org%2010.1371%20journal.pone.0235396.g001.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=258>) · Original paper and literature note are not hosted.
### P198 - Fig. 2
**Paper:** Validation of survey effort measures of grit and self-control in a sample of high school students
**Original caption:** Fig 2. Distribution of careless answering measures. https://doi.org/10.1371/journal.pone.0235396.g002
**Purpose:** Describes the careless-answering effort measure under validation. **Visual forms:** Histogram.
**Physical source pages:** 6.
[PNG](<04_Measurement_and_Data_Quality/P198-Fig02-Distribution%20of%20careless%20answering%20measures.%20https%20doi.org%2010.1371%20journal.pone.0235396.g002.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=259>) · Original paper and literature note are not hosted.
### P201 - Fig. 1
**Paper:** A psychometrics of individual differences in experimental tasks
**Original caption:** Fig. 1 The effect of the number-of-trials-per-individual on sample effect sizes and sample reliability for Stroop and flanker data from Hedge et al. (2018). Each point is the mean over 100 different samples at the indicated sample size
**Purpose:** Demonstrates trial-count dependence of conventional effect and reliability estimates. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 3.
[PNG](<04_Measurement_and_Data_Quality/P201-Fig01-The%20effect%20of%20the%20number-of-trials-per-individual%20on%20sample%20effect%20sizes%20and%20sample%20reliability%20for%20Stroop%20and%20flanker%20data%20from%20Hedge%20et%20al.%20(2018).%20Each%20point....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=769>) · Original paper and literature note are not hosted.
### P201 - Fig. 3
**Paper:** A psychometrics of individual differences in experimental tasks
**Original caption:** Fig. 3 Test–retest reliability for Hedge et al.’s Stroop-task data set. The top row shows the analysis for one block per session; the bottom row shows the analysis for all blocks. a, b Scatter plots of individuals” sample effects show almost no relation for one block and a moderate one for all blocks. c, d Scatter plots of model estimates for individuals shows a modest relationship for one block and a strong one for all blocks. e, d Posterior distribution of the model-based correlation parameter ρ. The correlation is modest and not well localized with one block; it is strong and well localized with all blocks
**Purpose:** Compares observed versus latent retest relations and uncertainty. **Visual forms:** Scatter, Density/histogram/boxplot, Composite.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P201-Fig03-Test-retest%20reliability%20for%20Hedge%20et%20al.%E2%80%99s%20Stroop-task%20data%20set.%20The%20top%20row%20shows%20the%20analysis%20for%20one%20block%20per%20session;%20the%20bottom%20row%20shows%20the%20analysis%20for....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=770>) · Original paper and literature note are not hosted.
### P205 - Table U01
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** 3.1 Action Library
**Purpose:** Original unnumbered hierarchical action-definition coding table; all three physical pages joined, with section heading retained and no invented table number. **Visual forms:** Complex table.
**Physical source pages:** 29, 30, 31.
[PNG](<04_Measurement_and_Data_Quality/P205-TableU01-3.1%20Action%20Library.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=771>) · Original paper and literature note are not hosted.
### P205 - Table U02
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** 3.2 Process Library
**Purpose:** Original unnumbered multilevel measurement codebook maps regulatory processes to definitions and sequential action patterns; full four-page table retained. **Visual forms:** Complex table.
**Physical source pages:** 31, 32, 33, 34.
[PNG](<04_Measurement_and_Data_Quality/P205-TableU02-3.2%20Process%20Library.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=772>) · Original paper and literature note are not hosted.
### P206 - Fig. 2
**Paper:** Beyond traditional interviews: Psychometric analysis of asynchronous video interviews for personality and interview performance evaluation using machine learning
**Original caption:** Fig. 2. Types of AI-based predicted scores. Notes. Green-shaded slices represent algorithmic bias absence; red-shaded slices represent algorithmic bias presence; ‘actual differences’ = actual group-differences (rater biases) in the observed ‘human-based’ data; for the quantitative comparison we applied a repeated measures general linear model (see also Table S1).
**Purpose:** Summarizes bias reproduction, amplification and attenuation in predicted assessments. **Visual forms:** Composite.
**Physical source pages:** 12.
[PNG](<04_Measurement_and_Data_Quality/P206-Fig02-Types%20of%20AI-based%20predicted%20scores.%20Notes.%20Green-shaded%20slices%20represent%20algorithmic%20bias%20absence;%20red-shaded%20slices%20represent%20algorithmic%20bias%20presence;%20%E2%80%98actual....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=773>) · Original paper and literature note are not hosted.
### P207 - Table U01
**Paper:** ChatGPT improves creative problem-solving performance in university students: An experimental study
**Original caption:** Appendix 1. Evaluation Matrix for a Product Improvement Task. Note. Evaluation criteria are adapted from Todd et al. (2019).
**Purpose:** Distinctive anchored scoring rubric compares five performance levels across quality, elaboration and originality, with worked examples; original has no table number. **Visual forms:** Complex table.
**Physical source pages:** 13.
[PNG](<04_Measurement_and_Data_Quality/P207-TableU01-Appendix%201.%20Evaluation%20Matrix%20for%20a%20Product%20Improvement%20Task.%20Note.%20Evaluation%20criteria%20are%20adapted%20from%20Todd%20et%20al.%20(2019).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=774>) · Original paper and literature note are not hosted.
### P212 - Fig. 2
**Paper:** Computer-based personality judgments are more accurate than those made by humans
**Original caption:** Fig. 2. Computer-based personality judgment accuracy (y axis), plotted against the number of Likes available for prediction (x axis). The red line represents the average accuracy (correlation) of computers’ judgment across the five personality traits. The five-trait average accuracy of human judgments is positioned onto the computer accuracy curve. For example, the accuracy of an average human individual (r = 0.49) is matched by that of the computer models based on around 90–100 Likes. The computer accuracy curves are smoothed using a LOWESS approach. The gray ribbon represents the 95% CI. Accuracy was averaged using Fisher’s r-to-z transformation.
**Purpose:** Validates computer personality judgments against self-ratings or external criteria. **Visual forms:** Line/profile plot.
**Physical source pages:** 3.
[PNG](<04_Measurement_and_Data_Quality/P212-Fig02-Computer-based%20personality%20judgment%20accuracy%20(y%20axis),%20plotted%20against%20the%20number%20of%20Likes%20available%20for%20prediction%20(x%20axis).%20The%20red%20line%20represents%20the%20average....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=775>) · Original paper and literature note are not hosted.
### P212 - Fig. 3
**Paper:** Computer-based personality judgments are more accurate than those made by humans
**Original caption:** Fig. 3. The external validity of personality judgments and self-ratings across the range of life outcomes, expressed as correlation (continuous variables; Upper) or AUC (dichotomous variables; Lower). The red, yellow, and blue bars indicate the external validity of self-ratings, human judgments, and computer judgments, respectively. For example, self-rated scores allow predicting network size with accuracy of r = 0.23, human judgments achieve r = 0.17 accuracy (or 0.06 less than self-ratings), whereas computer-based judgments achieve r = 0.24 accuracy (or 0.01 more than self-ratings). Compound variables (i.e., variables representing accuracy averaged across a few subvariables) are marked with an asterisk; see Table S4 for detailed results. Results are ordered by computer accuracy.
**Purpose:** Validates computer personality judgments against self-ratings or external criteria. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 4.
[PNG](<04_Measurement_and_Data_Quality/P212-Fig03-The%20external%20validity%20of%20personality%20judgments%20and%20self-ratings%20across%20the%20range%20of%20life%20outcomes,%20expressed%20as%20correlation%20(continuous%20variables;%20Upper)%20or%20AUC....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=776>) · Original paper and literature note are not hosted.
### P214 - Table 1
**Paper:** Developing measures of engagement in PISA
**Original caption:** Table 1. Measures of disengagement by category across PISA cycles
**Purpose:** Structured taxonomy and cycle-availability matrix of alternative engagement indicators. **Visual forms:** Complex table, Heatmap/matrix.
**Physical source pages:** 16.
[PNG](<04_Measurement_and_Data_Quality/P214-Table01-Measures%20of%20disengagement%20by%20category%20across%20PISA%20cycles.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=777>) · Original paper and literature note are not hosted.
### P214 - Table 3
**Paper:** Developing measures of engagement in PISA
**Original caption:** Table 3. Consistency among measures of disengagement (country level). Pearson correlation coefficients (upper triangular), scatter plots (lower triangular). Notes: The analyses are based on data from PISA 2018. All indices are coded so that higher values indicate higher levels of disengagement. *p < .05, †p < .10. Points in darker colour represent OECD countries.
**Purpose:** Triangular scatterplot/correlation matrix evaluates convergent evidence among disengagement measures. **Visual forms:** Scatter, Heatmap/matrix, Complex table.
**Physical source pages:** 23, 24.
[PNG](<04_Measurement_and_Data_Quality/P214-Table03-Consistency%20among%20measures%20of%20disengagement%20(country%20level).%20Pearson%20correlation%20coefficients%20(upper%20triangular),%20scatter%20plots%20(lower%20triangular).%20Notes%20The....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=778>) · Original paper and literature note are not hosted.
### P214 - Fig. 5
**Paper:** Developing measures of engagement in PISA
**Original caption:** Figure 5. Relationship between disengagement measures and reading performance. Notes: The analyses are based on data from PISA 2018. Values represent the predicted change on the PISA Reading scale associated with a one standard deviation increase on the respective disengagement measure after accounting for student characteristics (e.g. gender and socio-economic status). Rapid guessing is only available for countries/economies that administered computer-based assessment.
**Purpose:** Criterion-related evidence connecting the alternative engagement indicators to reading achievement. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 30.
[PNG](<04_Measurement_and_Data_Quality/P214-Fig05-Relationship%20between%20disengagement%20measures%20and%20reading%20performance.%20Notes%20The%20analyses%20are%20based%20on%20data%20from%20PISA%202018.%20Values%20represent%20the%20predicted%20change%20on....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=779>) · Original paper and literature note are not hosted.
### P214 - Fig. 6
**Paper:** Developing measures of engagement in PISA
**Original caption:** Figure 6. Association of self-reported effort in PISA 2018 and 2012. Note: Points represent country means; points in darker colour represent OECD countries.
**Purpose:** Cross-cycle country-mean stability evidence for the disengagement indicator. **Visual forms:** Scatter.
**Physical source pages:** 32.
[PNG](<04_Measurement_and_Data_Quality/P214-Fig06-Association%20of%20self-reported%20effort%20in%20PISA%202018%20and%202012.%20Note%20Points%20represent%20country%20means;%20points%20in%20darker%20colour%20represent%20OECD%20countries.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=780>) · Original paper and literature note are not hosted.
### P214 - Fig. 7
**Paper:** Developing measures of engagement in PISA
**Original caption:** Figure 7. Association of non-response in PISA 2018, 2015 and 2012. Notes: Points represent country means; points in darker colour represent OECD countries.
**Purpose:** Cross-cycle country-mean stability evidence for the disengagement indicator. **Visual forms:** Scatter.
**Physical source pages:** 32.
[PNG](<04_Measurement_and_Data_Quality/P214-Fig07-Association%20of%20non-response%20in%20PISA%202018,%202015%20and%202012.%20Notes%20Points%20represent%20country%20means;%20points%20in%20darker%20colour%20represent%20OECD%20countries.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=781>) · Original paper and literature note are not hosted.
### P214 - Fig. 8
**Paper:** Developing measures of engagement in PISA
**Original caption:** Figure 8. Association of rapid guessing in PISA 2018 and 2015 for Reading, Science, and Mathematics. Note: Points represent country means; points in darker colour represent OECD countries.
**Purpose:** Cross-cycle country-mean stability evidence for the disengagement indicator. **Visual forms:** Scatter.
**Physical source pages:** 33.
[PNG](<04_Measurement_and_Data_Quality/P214-Fig08-Association%20of%20rapid%20guessing%20in%20PISA%202018%20and%202015%20for%20Reading,%20Science,%20and%20Mathematics.%20Note%20Points%20represent%20country%20means;%20points%20in%20darker%20colour%20represent....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=782>) · Original paper and literature note are not hosted.
### P216 - Table 1
**Paper:** Does geographical location impact educational disparities among Ecuadorians? A novel two-stage inequality decomposition method
**Original caption:** Table 1. Procedure for estimating the variable ‘years of schooling’ (2014 and 2021). NOTES: y = number of years of the answer; SOURCE: Authors ́ own based on the methodology proposed by the Ministerio de Educación (2015). aFormer Education System (1983 Ley Orgánica de Educación [Education Act]); bCurrent Education System (2011 Ley Orgánica de Educación Intercultural [LOEI, or Intercultural Education Act], amended in 2021 by the Ley Orgánica Reformatoria de la Ley Orgánica de Educación Intercultural [LOEI Reform Act]; For individuals trained in previous education systems, the Ecuador’s National Survey of Employment, Unemployment and Underemployment considers the number of years of schooling to be equal to 2 y, when y = 0, 1, 2 and 3 + y, when y = 3, 4, 5, 6, 7, 8, 9, 10.
**Purpose:** Explicit mapping between former/current education systems and derived years of schooling, including methodological notes. **Visual forms:** Complex table, Measurement conversion.
**Physical source pages:** 6.
[PNG](<04_Measurement_and_Data_Quality/P216-Table01-Procedure%20for%20estimating%20the%20variable%20%E2%80%98years%20of%20schooling%E2%80%99%20(2014%20and%202021).%20NOTES%20y%20=%20number%20of%20years%20of%20the%20answer;%20SOURCE%20Authors%20%CC%81%20own%20based%20on%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=783>) · Original paper and literature note are not hosted.
### P216 - App Table A2
**Paper:** Does geographical location impact educational disparities among Ecuadorians? A novel two-stage inequality decomposition method
**Original caption:** TABLE A2 Latest Education Systems in Ecuador (non-university education). SOURCE: Authors’ own based on 1983 Ley Orgánica de Educación, or Education Act and 2011 Ley Orgánica de Educación Intercultural [LOEI, or Intercultural Education Act], amended in 2021 by the Ley Orgánica Reformatoria de la Ley Orgánica de Educación Intercultural [LOEI Reform Act].
**Purpose:** Merged two-system comparison maps Ecuadorian schooling stages under the former and current education laws; a distinctive measurement crosswalk. **Visual forms:** Complex table, Classification crosswalk.
**Physical source pages:** 2.
[PNG](<04_Measurement_and_Data_Quality/P216-App-TableA2-Latest%20Education%20Systems%20in%20Ecuador%20(non-university%20education).%20SOURCE%20Authors%E2%80%99%20own%20based%20on%201983%20Ley%20Org%C3%A1nica%20de%20Educaci%C3%B3n,%20or%20Education%20Act%20and%202011%20Ley....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=784>) · Original paper and literature note are not hosted.
### P217 - Fig. 1
**Paper:** Do smartphone usage scales predict behavior?
**Original caption:** Figure 1. Average r value for each subjective measure across all three objective behavioral measures. Error bars illustrate standard error. Red indicates a single behavioral estimate. Dotted line represents mean correlation across all measures. Refer to Table 1 for abbreviations.
**Purpose:** Compares behavioral validity of subjective phone-use measures. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 15.
[PNG](<04_Measurement_and_Data_Quality/P217-Fig01-Average%20r%20value%20for%20each%20subjective%20measure%20across%20all%20three%20objective%20behavioral%20measures.%20Error%20bars%20illustrate%20standard%20error.%20Red%20indicates%20a%20single....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=785>) · Original paper and literature note are not hosted.
### P218 - Fig. 2
**Paper:** Estimating Measurement Quality in Digital Trace Data and Surveys Using the MultiTrait MultiMethod Model
**Original caption:** Figure 2. Correlation matrix of the five traits (calling, messaging, taking photos, social media usage, and web browsing) and the five traits (5-point survey, 7-point survey, duration survey, count digital trace data, duration digital trace data).
**Purpose:** Compares measurement quality across survey and trace methods. **Visual forms:** Heatmap/matrix.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P218-Fig02-Correlation%20matrix%20of%20the%20five%20traits%20(calling,%20messaging,%20taking%20photos,%20social%20media%20usage,%20and%20web%20browsing)%20and%20the%20five%20traits%20(5-point%20survey,%207-point....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=786>) · Original paper and literature note are not hosted.
### P218 - Fig. 3
**Paper:** Estimating Measurement Quality in Digital Trace Data and Surveys Using the MultiTrait MultiMethod Model
**Original caption:** Figure 3. Variance decomposition over all the questions by method based on the MTMM.
**Purpose:** Compares measurement quality across survey and trace methods. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 9.
[PNG](<04_Measurement_and_Data_Quality/P218-Fig03-Variance%20decomposition%20over%20all%20the%20questions%20by%20method%20based%20on%20the%20MTMM.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=787>) · Original paper and literature note are not hosted.
### P218 - Fig. 4
**Paper:** Estimating Measurement Quality in Digital Trace Data and Surveys Using the MultiTrait MultiMethod Model
**Original caption:** Figure 4. Results from the MTMM model. Trait variance represent “valid” measurement while method and random error are two different sources of measurement error.
**Purpose:** Compares measurement quality across survey and trace methods. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 9.
[PNG](<04_Measurement_and_Data_Quality/P218-Fig04-Results%20from%20the%20MTMM%20model.%20Trait%20variance%20represent%20%E2%80%9Cvalid%E2%80%9D%20measurement%20while%20method%20and%20random%20error%20are%20two%20different%20sources%20of%20measurement%20error.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=788>) · Original paper and literature note are not hosted.
### P219 - Fig. 3
**Paper:** Explaining social disparities in ICT literacy: the role of multiple socioeconomic status indicators and ICT-related motivations
**Original caption:** Fig. 3 Higher order factor CFA model of the ICT usage motives. Inf=information seeking, LW=learn and work, SEX=social exchange, SP=self-presentation
**Purpose:** Higher-order CFA measurement structure and factor loadings, not a structural outcome model. **Visual forms:** CFA path diagram.
**Physical source pages:** 14.
[PNG](<04_Measurement_and_Data_Quality/P219-Fig03-Higher%20order%20factor%20CFA%20model%20of%20the%20ICT%20usage%20motives.%20Inf=information%20seeking,%20LW=learn%20and%20work,%20SEX=social%20exchange,%20SP=self-presentation.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=789>) · Original paper and literature note are not hosted.
### P221 - Fig. 4
**Paper:** From asking to observing. Behavioural measures of socio-emotional and motivational skills in large-scale assessments
**Original caption:** Fig. 4. Associations between PSEC indicators and mathematics achievement. Notes: An asterisk (*) next to an indicator name indicates that the estimated coefficient is statistically significant at least at the 5% level. A dark shade indicates that the estimated association for a specific country is statistically significantly above the average for that indicator across the countries with available information (p ≤ 0.05). Data for all countries in the sample can be consulted in Table K1 in the Online Supplementary Annex and data for other PISA outcomes can be consulted in Tables K2-K4. Source: PISA 2012data.
**Purpose:** Validates PSEC indicators against mathematics achievement. **Visual forms:** Bar chart.
**Physical source pages:** 15.
[PNG](<04_Measurement_and_Data_Quality/P221-Fig04-Associations%20between%20PSEC%20indicators%20and%20mathematics%20achievement.%20Notes%20An%20asterisk%20(%201%20asterisk%20)%20next%20to%20an%20indicator%20name%20indicates%20that%20the%20estimated....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=790>) · Original paper and literature note are not hosted.
### P226 - Table 2
**Paper:** Granularity matters: comparing different ways of measuring self-regulated learning
**Original caption:** Table 2 Schematic overview of studies comparing specific self-regulatory strategies, with + indicating high calibration, − indicating low calibration, and +/− indicating mixed results
**Purpose:** Structured comparison of study methods, calibration evidence, tasks and achievement prediction, central to the granularity argument. **Visual forms:** Complex table.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P226-Table02-Schematic%20overview%20of%20studies%20comparing%20specific%20self-regulatory%20strategies,%20with%20+%20indicating%20high%20calibration,%20-%20indicating%20low%20calibration,%20and%20+....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=791>) · Original paper and literature note are not hosted.
### P226 - Table 3
**Paper:** Granularity matters: comparing different ways of measuring self-regulated learning
**Original caption:** Table 3 Schematic overview of studies comparing global strategy use, with + indicating high calibration, − indicating low calibration, and +/− indicating mixed results
**Purpose:** Structured comparison of study methods, calibration evidence, tasks and achievement prediction, central to the granularity argument. **Visual forms:** Complex table.
**Physical source pages:** 9.
[PNG](<04_Measurement_and_Data_Quality/P226-Table03-Schematic%20overview%20of%20studies%20comparing%20global%20strategy%20use,%20with%20+%20indicating%20high%20calibration,%20-%20indicating%20low%20calibration,%20and%20+%20-%20indicating%20mixed%20results.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=792>) · Original paper and literature note are not hosted.
### P228 - Fig. 7
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Fig. 7 Adjacency matrix of transitions between states and percent missing values for total time on state the variables
**Purpose:** Diagnoses extracted timing indicators, structural missingness or response-time relationships. **Visual forms:** Composite.
**Physical source pages:** 24.
[PNG](<04_Measurement_and_Data_Quality/P228-Fig07-Adjacency%20matrix%20of%20transitions%20between%20states%20and%20percent%20missing%20values%20for%20total%20time%20on%20state%20the%20variables.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=793>) · Original paper and literature note are not hosted.
### P228 - Fig. 8
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Fig. 8 Average of time components (total time on state) and frequency of item-level response times for the analyzed item
**Purpose:** Diagnoses extracted timing indicators, structural missingness or response-time relationships. **Visual forms:** Composite.
**Physical source pages:** 25.
[PNG](<04_Measurement_and_Data_Quality/P228-Fig08-Average%20of%20time%20components%20(total%20time%20on%20state)%20and%20frequency%20of%20item-level%20response%20times%20for%20the%20analyzed%20item.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=794>) · Original paper and literature note are not hosted.
### P228 - Fig. 9
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Fig. 9 Total time on state “Stem_Q01”, “Stem_Q02” and “Stem_Q3” conditional the response to the first question Q01
**Purpose:** Diagnoses extracted timing indicators, structural missingness or response-time relationships. **Visual forms:** Composite.
**Physical source pages:** 26.
[PNG](<04_Measurement_and_Data_Quality/P228-Fig09-Total%20time%20on%20state%20%E2%80%9CStem_Q01%E2%80%9D,%20%E2%80%9CStem_Q02%E2%80%9D%20and%20%E2%80%9CStem_Q3%E2%80%9D%20conditional%20the%20response%20to%20the%20first%20question%20Q01.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=795>) · Original paper and literature note are not hosted.
### P228 - Fig. 10
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Fig. 10 Total time on state “Q01”, “Q02”, …, “Q13” conditional the response the question (item-level response after
**Purpose:** Diagnoses extracted timing indicators, structural missingness or response-time relationships. **Visual forms:** Composite.
**Physical source pages:** 27.
[PNG](<04_Measurement_and_Data_Quality/P228-Fig10-Total%20time%20on%20state%20%E2%80%9CQ01%E2%80%9D,%20%E2%80%9CQ02%E2%80%9D,%20...,%20%E2%80%9CQ13%E2%80%9D%20conditional%20the%20response%20the%20question%20(item-level%20response%20after.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=796>) · Original paper and literature note are not hosted.
### P234 - Fig. 2
**Paper:** Large studies reveal how reference bias limits policy applications of self-report measures
**Original caption:** Figure 2. In Study 1, self-reported grit correlated positively with a student’s own academic performance but inversely with the performance of their schoolmates. OLS models included demographic controls and school fixed effects. Error bars represent 95% confidence intervals. Model R2s for GPA, math score, and language score were 0.124, 0.071, and 0.071, respectively.
**Purpose:** Tests reference bias in self-report measurement at within- and between-group levels. **Visual forms:** Bar chart.
**Physical source pages:** 4.
[PNG](<04_Measurement_and_Data_Quality/P234-Fig02-In%20Study%201,%20self-reported%20grit%20correlated%20positively%20with%20a%20student%E2%80%99s%20own%20academic%20performance%20but%20inversely%20with%20the%20performance%20of%20their%20schoolmates.%20OLS%20models....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=797>) · Original paper and literature note are not hosted.
### P234 - Fig. 3
**Paper:** Large studies reveal how reference bias limits policy applications of self-report measures
**Original caption:** Figure 3. In Study 2, self-reported conscientiousness correlated positively with a student’s own GPA and negatively with the GPA of near-peers. In contrast, standards for what constitutes hard work and preparedness correlated positively with both own and near-peer GPA. As expected, there was no effect of far-peer GPA. OLS models included demographic controls and school fixed effects. Error bars represent 95% confidence intervals. Model R2s for conscientiousness, hard work norms, and preparedness norms were 0.095, 0.159, and 0.059, respectively.
**Purpose:** Tests reference bias in self-report measurement at within- and between-group levels. **Visual forms:** Bar chart.
**Physical source pages:** 6.
[PNG](<04_Measurement_and_Data_Quality/P234-Fig03-In%20Study%202,%20self-reported%20conscientiousness%20correlated%20positively%20with%20a%20student%E2%80%99s%20own%20GPA%20and%20negatively%20with%20the%20GPA%20of%20near-peers.%20In%20contrast,%20standards%20for....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=798>) · Original paper and literature note are not hosted.
### P234 - Fig. 4
**Paper:** Large studies reveal how reference bias limits policy applications of self-report measures
**Original caption:** Figure 4. In Study 3, comparing students within schools (colored lines), higher self-regulation predicted higher odds of college graduation, whether measured by self-report questionnaires for grit and self-control or by a behavioral task called the Academic Diligence Task. When comparing schools to each other, however, higher self-reported grit and self-control scores predicted lower graduation rates, whereas the behavioral task positively predicted college graduation, as shown in the solid black lines. Plots show predicted probabilities of graduation from multilevel logistic regression models. AUCs for models predicting the academic diligence task, self-reported grit, and self-reported self-control were 0.694, 0.693, and, 0.676, respectively.
**Purpose:** Tests reference bias in self-report measurement at within- and between-group levels. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 7.
[PNG](<04_Measurement_and_Data_Quality/P234-Fig04-In%20Study%203,%20comparing%20students%20within%20schools%20(colored%20lines),%20higher%20self-regulation%20predicted%20higher%20odds%20of%20college%20graduation,%20whether%20measured%20by%20self-report....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=799>) · Original paper and literature note are not hosted.
### P236 - Fig. 1
**Paper:** Measuring Test-Taking Effort on Constructed-Response Items with Item Response Time and Number of Actions
**Original caption:** Figure 1. Clustered boxplots of accuracy level for 44 items by effort measure and by response behaviour
**Purpose:** Validates four item-effort flags through accuracy or missingness. **Visual forms:** Density/histogram/boxplot, Composite.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P236-Fig01-Clustered%20boxplots%20of%20accuracy%20level%20for%2044%20items%20by%20effort%20measure%20and%20by%20response%20behaviour.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=800>) · Original paper and literature note are not hosted.
### P236 - Fig. 2
**Paper:** Measuring Test-Taking Effort on Constructed-Response Items with Item Response Time and Number of Actions
**Original caption:** Figure 2. Clustered boxplots of missingness proportions for 44 items by effort measure and by response behaviour
**Purpose:** Validates four item-effort flags through accuracy or missingness. **Visual forms:** Density/histogram/boxplot, Composite.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P236-Fig02-Clustered%20boxplots%20of%20missingness%20proportions%20for%2044%20items%20by%20effort%20measure%20and%20by%20response%20behaviour.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=801>) · Original paper and literature note are not hosted.
### P239 - Fig. 1
**Paper:** Reliability of educational attainment of survey respondents: an overlooked barrier to comparability?
**Original caption:** Fig. 1 Operationalization of educational variables for comparability. Adapted from Schneider (2009, p. 57).
**Purpose:** Operationalization framework links constructs, educational-attainment questions and harmonized classifications. **Visual forms:** Flowchart, Measurement framework.
**Physical source pages:** 3.
[PNG](<04_Measurement_and_Data_Quality/P239-Fig01-Operationalization%20of%20educational%20variables%20for%20comparability.%20Adapted%20from%20Schneider%20(2009,%20p.%2057).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=802>) · Original paper and literature note are not hosted.
### P239 - Fig. 2
**Paper:** Reliability of educational attainment of survey respondents: an overlooked barrier to comparability?
**Original caption:** Fig. 2 Flow of inconsistencies between t1 and t2 in percentages for measurement of education harmonised to ISCED 2011 1-digit (levels of education) Description of ISCED 2011 1-digit codes: 0 = Less than primary education; 1 = Primary education; 2 = Lower secondary education; 3 = Upper secondary education; 4 = Post-secondary non-tertiary education; 5 = Short-cycle tertiary education; 6 = Bachelor’s or equivalent level; 7 = Master’s or equivalent level; 8 = Doctoral or equivalent level.
**Purpose:** Test–retest inconsistency flows assess reliability of harmonized educational-attainment classifications; all three country panels retained. **Visual forms:** Alluvial diagram, Composite.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P239-Fig02-Flow%20of%20inconsistencies%20between%20t1%20and%20t2%20in%20percentages%20for%20measurement%20of%20education%20harmonised%20to%20ISCED%202011%201-digit%20(levels%20of%20education)%20Description%20of%20ISCED....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=803>) · Original paper and literature note are not hosted.
### P239 - Fig. 3
**Paper:** Reliability of educational attainment of survey respondents: an overlooked barrier to comparability?
**Original caption:** Fig. 3 Flow of inconsistencies between t1 and t2 in percentages for measurement of education harmonised to EDULVLB (detailed measurement). Description of EDULVLB codes with reference to ISCED-97: 000 = Not completed ISCED level 1; 113 = ISCED 1, completed primary education; 129 = Vocational ISCED 2C < 2 years, no access ISCED 3; 212 = General/pre-vocational ISCED 2A/2B, access ISCED 3 vocational; 213 = General ISCED 2A, access ISCED 3A general/all 3; 221 = Vocational ISCED 2C ≥ 2 years, no access ISCED 3; 222 = Vocational ISCED 2A/2B, access ISCED 3 vocational; 223 = Vocational ISCED 2, access ISCED 3 general/all; 229 = Vocational ISCED 3C < 2 years, no access ISCED 5; 311 = General ISCED 3 ≥ 2 years, no access ISCED 5; 312 = General ISCED 3A/3B, access ISCED 5B/lower tier 5A; 313 = General ISCED 3A, access upper tier ISCED 5A/all 5; 321 = Vocational ISCED 3C ≥ 2 years, no access ISCED 5; 323 = Vocational ISCED 3A, access upper tier ISCED 5A/all 5; 412 = General ISCED 4A/4B, access ISCED 5B/ lower tier 5A; 413 = General ISCED 4A, access upper tier ISCED 5A/all 5; 421 = ISCED 4 programmes without access ISCED 5; 422 = Vocational ISCED 4A/4B, access ISCED 5B/lower tier 5A; 423 = Vocational ISCED 4A, access upper tier ISCED 5A/all 5; 510 = ISCED 5A short, intermediate/academic/ general tertiary below; 520 = ISCED 5B short, advanced vocational qualifications; 610 = ISCED 5A medium, bachelor/equivalent from lower tier tertiary; 620 = ISCED 5A medium, bachelor/equivalent from upper/single tier; 710 = ISCED 5A long, master/equivalent from lower tier tertiary; 720 = ISCED 5A long, master/equivalent from upper/single tier tertiary; 800 = ISCED 6, doctoral degree.
**Purpose:** Test–retest inconsistency flows assess reliability of harmonized educational-attainment classifications; all three country panels retained. **Visual forms:** Alluvial diagram, Composite.
**Physical source pages:** 9.
[PNG](<04_Measurement_and_Data_Quality/P239-Fig03-Flow%20of%20inconsistencies%20between%20t1%20and%20t2%20in%20percentages%20for%20measurement%20of%20education%20harmonised%20to%20EDULVLB%20(detailed%20measurement).%20Description%20of%20EDULVLB%20codes....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=804>) · Original paper and literature note are not hosted.
### P239 - Fig. 4
**Paper:** Reliability of educational attainment of survey respondents: an overlooked barrier to comparability?
**Original caption:** Fig. 4 Effect of device type on response consistency of educational measurement t1-t2 across countries. Use of smartphones equals 1; consistency of answers between t1 and t2 equals 1.
**Purpose:** Device-related response-consistency differences diagnose survey measurement quality; ordinary Table 3 above is excluded. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 10.
[PNG](<04_Measurement_and_Data_Quality/P239-Fig04-Effect%20of%20device%20type%20on%20response%20consistency%20of%20educational%20measurement%20t1-t2%20across%20countries.%20Use%20of%20smartphones%20equals%201;%20consistency%20of%20answers%20between%20t1....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=805>) · Original paper and literature note are not hosted.
### P240 - Fig. 1
**Paper:** Response Time Effort: A New Measure of Examinee Motivation in Computer-Based Tests
**Original caption:** Figure 1. Response time frequency polygons for two ILT items.
**Purpose:** Validation of response-time effort and item-level disengagement. **Visual forms:** Distribution, Measurement diagnostic.
**Physical source pages:** 21, 22.
[PNG](<04_Measurement_and_Data_Quality/P240-Fig01-Response%20time%20frequency%20polygons%20for%20two%20ILT%20items.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=806>) · Original paper and literature note are not hosted.
### P240 - Fig. 2
**Paper:** Response Time Effort: A New Measure of Examinee Motivation in Computer-Based Tests
**Original caption:** Figure 2. Histogram of Response Time Effort (RTE) scores.
**Purpose:** Validation of response-time effort and item-level disengagement. **Visual forms:** Distribution, Measurement diagnostic.
**Physical source pages:** 21, 23.
[PNG](<04_Measurement_and_Data_Quality/P240-Fig02-Histogram%20of%20Response%20Time%20Effort%20(RTE)%20scores.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=807>) · Original paper and literature note are not hosted.
### P240 - Fig. 3
**Paper:** Response Time Effort: A New Measure of Examinee Motivation in Computer-Based Tests
**Original caption:** Figure 3. Scatter plot of RTE scores by self-reported effort.
**Purpose:** Recovered by direct visual comparison with source page. **Visual forms:** Scatter plot, Measurement diagnostic.
**Physical source pages:** 21, 23.
[PNG](<04_Measurement_and_Data_Quality/P240-Fig03-Scatter%20plot%20of%20RTE%20scores%20by%20self-reported%20effort.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=808>) · Original paper and literature note are not hosted.
### P240 - Fig. 4
**Paper:** Response Time Effort: A New Measure of Examinee Motivation in Computer-Based Tests
**Original caption:** Figure 4. Distribution of item response times by item position for an examinee who appeared to switch response strategies in the middle of the test.
**Purpose:** Validation of response-time effort and item-level disengagement. **Visual forms:** Scatter plot, Measurement diagnostic.
**Physical source pages:** 21, 24.
[PNG](<04_Measurement_and_Data_Quality/P240-Fig04-Distribution%20of%20item%20response%20times%20by%20item%20position%20for%20an%20examinee%20who%20appeared%20to%20switch%20response%20strategies%20in%20the%20middle%20of%20the%20test.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=809>) · Original paper and literature note are not hosted.
### P243 - Fig. 1
**Paper:** Socioeconomic gradients in children’s cognitive skills: are cross-country comparisons robust to who reports family background?
**Original caption:** Figure 1. Difference in average test scores between children reporting the same SES category as their parents and those who do not (PISA and PIRLS 2006). Notes: The graph shows the difference in average reading test scores between children who report the same SES category as their parents and those who do not (subtracting the latter from the former). All figures refer to national standard deviations. The thin line running through the country name shows the estimated 95% confidence interval. Final student weights applied. Results for occupation and education refer to PISA. Results for books in the home refer to PIRLS.
**Purpose:** Reporter-disagreement diagnostic across countries and three family-background measures. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 27.
[PNG](<04_Measurement_and_Data_Quality/P243-Fig01-Difference%20in%20average%20test%20scores%20between%20children%20reporting%20the%20same%20SES%20category%20as%20their%20parents%20and%20those%20who%20do%20not%20(PISA%20and%20PIRLS%202006).%20Notes%20The%20graph....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=810>) · Original paper and literature note are not hosted.
### P250 - Table 1
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Table 1. Changes to the question order in 2022
**Purpose:** Structured cross-cycle comparison of questionnaire order, relevant to measurement continuity. **Visual forms:** Complex table, Instrument comparison.
**Physical source pages:** 10.
[PNG](<04_Measurement_and_Data_Quality/P250-Table01-Changes%20to%20the%20question%20order%20in%202022.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=811>) · Original paper and literature note are not hosted.
### P250 - Fig. 1
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 1. Percentage of 15-year-old students with missing occupational status scores Notes: A consistent conversion of occupational codes to occupational status scores (ISEI) was applied for all years. Average rates of missingness across 59 countries and economies are shown on the left. Each line in the right-hand chart represents the rate of missingness of the “highest” ISEI score for a distinct country/economy; only the most remarkable patterns are labeled with three-letter ISO codes. Source: PISA 2015, 2018 and 2022 Datasets, available at OECD (2024[7]), PISA Data and Methodology, https://www.oecd.org/en/about/programmes/pisa/pisa-data.html.
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 15.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig01-Percentage%20of%2015-year-old%20students%20with%20missing%20occupational%20status%20scores%20Notes%20A%20consistent%20conversion%20of%20occupational%20codes%20to%20occupational%20status%20scores....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=812>) · Original paper and literature note are not hosted.
### P250 - Table 2
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Table 2. Changes to parental education questions in 2018 and 2022
**Purpose:** Side-by-side cross-cycle parental-education question changes. **Visual forms:** Complex table, Instrument comparison.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P250-Table02-Changes%20to%20parental%20education%20questions%20in%202018%20and%202022.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=813>) · Original paper and literature note are not hosted.
### P250 - Fig. 2
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 2. Missing and invalid reports about parents’ occupation Percentage of 15-year-old students (average across 59 trend countries and economies) Note: A consistent conversion of occupational codes to occupational status scores (ISEI) was applied for all years. Numbers inside the figure correspond to the share of students in each category. Source: PISA 2015, 2018 and 2022 Datasets, , available at OECD (2024[7]), PISA Data and Methodology, https://www.oecd.org/en/about/programmes/pisa/pisa-data.html.
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 16.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig02-Missing%20and%20invalid%20reports%20about%20parents%E2%80%99%20occupation%20Percentage%20of%2015-year-old%20students%20(average%20across%2059%20trend%20countries%20and%20economies)%20Note%20A%20consistent....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=814>) · Original paper and literature note are not hosted.
### P250 - Fig. 3
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 3. Occupational status scores and major ISCO groups among parents of 15-year-olds Average values across 59 trend countries and economies Note: A consistent conversion of occupational codes to occupational status scores (ISEI) was applied for all years. The blue-shaded areas in the right-hand figures represent major ISCO groups, ordered, from top to bottom, in ascending order of the digit which identifies them (numbers correspond to the share of students in each category): from major group “0” (Armed forces occupations) at the top to major group “9” (Elementary occupations) at the bottom. Intermediate categories correspond to Managers, senior officials and legislators (Major group 1); Professionals (Major group 2); Technicians and associate professionals (Major group 3); Clerks (Major group 4); Service and sales workers (Major group 5); Skilled agricultural and fishery workers (Major group 6); Craft and related trades workers (Major group 7); Plant and machine operators, and assemblers (Major group 8). The grey area at the bottom corresponds to inactive (e.g. “stay at home mom”), students, and social beneficiaries. Source: PISA 2015, 2018 and 2022 Datasets, , available at OECD (2024[7]), PISA Data and Methodology, https://www.oecd.org/en/about/programmes/pisa/pisa-data.html.
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 17.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig03-Occupational%20status%20scores%20and%20major%20ISCO%20groups%20among%20parents%20of%2015-year-olds%20Average%20values%20across%2059%20trend%20countries%20and%20economies%20Note%20A%20consistent%20conversion....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=815>) · Original paper and literature note are not hosted.
### P250 - Fig. 4
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 4. Missing occupational status scores, by decile of the index of household possession Average values across 59 trend countries and economies Source: PISA 2015, 2018 and 2022 Datasets, , available at OECD (2024[7]), PISA Data and Methodology, https://www.oecd.org/en/about/programmes/pisa/pisa-data.html.
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 18.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig04-Missing%20occupational%20status%20scores,%20by%20decile%20of%20the%20index%20of%20household%20possession%20Average%20values%20across%2059%20trend%20countries%20and%20economies%20Source%20PISA%202015,%202018....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=816>) · Original paper and literature note are not hosted.
### P250 - Fig. 5
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 5. Trends in parental education Average values across 59 trend countries and economies Note: A consistent conversion of educational qualifications to years of education (PAREDINT) was applied for all years. Source: PISA 2015, 2018 and 2022 Datasets, , available at OECD (2024[7]), PISA Data and Methodology, https://www.oecd.org/en/about/programmes/pisa/pisa-data.html.
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 20.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig05-Trends%20in%20parental%20education%20Average%20values%20across%2059%20trend%20countries%20and%20economies%20Note%20A%20consistent%20conversion%20of%20educational%20qualifications%20to%20years%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=817>) · Original paper and literature note are not hosted.
### P250 - Fig. 6
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 6. External and internal validity evidence for the share of tertiary-educated parents in PISA Note: The top row is for women (mothers, in PISA); the bottom row for men (fathers). The left-most chart shows a comparison of levels (in percentages) across independent data sources; the middle charts show a comparisons of trends (percentage-point differences) across independent data sources; the right-most charts show an comparison of percentage-point differences across adjacent cycles in PISA. Source: PISA 2015, 2018 and 2022 Datasets, available at OECD (2024[7]), PISA Data and Methodology, https://www.oecd.org/en/about/programmes/pisa/pisa-data.html; and OECD (2023[8]), Education at a Glance (EAG) Database, https://doi.org/10.1787/48e69087-en.
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 21.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig06-External%20and%20internal%20validity%20evidence%20for%20the%20share%20of%20tertiary-educated%20parents%20in%20PISA%20Note%20The%20top%20row%20is%20for%20women%20(mothers,%20in%20PISA);%20the%20bottom%20row%20for....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=818>) · Original paper and literature note are not hosted.
### P250 - Fig. 7
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 7. Trends in household possessions (trend items) Average values across 59 trend countries and economies Note: Response categories were merged for trend comparisons. Source: PISA 2015, 2018 and 2022 Datasets, available at OECD (2024[7]), PISA Data and Methodology, https://www.oecd.org/en/about/programmes/pisa/pisa-data.html
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 23.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig07-Trends%20in%20household%20possessions%20(trend%20items)%20Average%20values%20across%2059%20trend%20countries%20and%20economies%20Note%20Response%20categories%20were%20merged%20for%20trend%20comparisons....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=819>) · Original paper and literature note are not hosted.
### P250 - Fig. 8
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 8. Agreement, over- and underreporting of student reports relative to parent reports about parental education and occupation Note: Analyses based on unweighted, pooled data, from 14 countries/economies. Source: OECD (2022[9]), PISA Field-Trial datasets for 2020/2021.
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Bar chart, Measurement diagnostic.
**Physical source pages:** 26, 27.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig08-Agreement,%20over-%20and%20underreporting%20of%20student%20reports%20relative%20to%20parent%20reports%20about%20parental%20education%20and%20occupation%20Note%20Analyses%20based%20on%20unweighted,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=820>) · Original paper and literature note are not hosted.
### P250 - Fig. 9
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 9. Student-reported parental education, by parent report Conditional distribution of highest education among parents or guardians, based on parent and student reports, collapsed into three broad educational groups Note: Dark shades represent agreeing student and parent reports; lighter shades are used for discrepant student and parent reports. Source: OECD (2022[9]), PISA Field-Trial datasets for 2020/2021 (14 countries/economies)
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Bar chart, Measurement diagnostic.
**Physical source pages:** 27.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig09-Student-reported%20parental%20education,%20by%20parent%20report%20Conditional%20distribution%20of%20highest%20education%20among%20parents%20or%20guardians,%20based%20on%20parent%20and%20student....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=821>) · Original paper and literature note are not hosted.
### P250 - Fig. 10
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 10. Student-reported parental occupation, by parent report Conditional distribution of highest occupational status among parents or guardians, based on parent and student reports, collapsed into 1-digit major ISCO groups Note: Major ISCO groups are ordered based on the occupational status value assigned, from lower to higher. PISA-specific codes (housewife, student, welfare beneficiary) are collapsed into major ISCO group 9. Source OECD (2022[9]), PISA Field-Trial datasets for 2020/2021 (14 countries/economies).
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Bar chart, Measurement diagnostic.
**Physical source pages:** 29.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig10-Student-reported%20parental%20occupation,%20by%20parent%20report%20Conditional%20distribution%20of%20highest%20occupational%20status%20among%20parents%20or%20guardians,%20based%20on%20parent%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=822>) · Original paper and literature note are not hosted.
### P250 - Fig. 11
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 11. Proportion missing values in parental occupation under pre-coded and post-coded formats Source: OECD (2022[9]), PISA Field-Trial datasets for 2020/2021 (60 countries/economies).
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Bar chart, Measurement diagnostic.
**Physical source pages:** 30.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig11-Proportion%20missing%20values%20in%20parental%20occupation%20under%20pre-coded%20and%20post-coded%20formats%20Source%20OECD%20(2022%5B9%5D),%20PISA%20Field-Trial%20datasets%20for%202020%202021%20(60....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=823>) · Original paper and literature note are not hosted.
### P250 - Fig. 12
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 12. Agreement and disagreement between response formats for parental occupation Based on student responses, pooling data for all parents and guardians Source: OECD (2022[9]), PISA Field-Trial datasets for 2020/2021 (60 countries/economies).
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Sankey/alluvial diagram, Measurement diagnostic.
**Physical source pages:** 32.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig12-Agreement%20and%20disagreement%20between%20response%20formats%20for%20parental%20occupation%20Based%20on%20student%20responses,%20pooling%20data%20for%20all%20parents%20and%20guardians%20Source%20OECD....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=824>) · Original paper and literature note are not hosted.
### P250 - Fig. 13
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 13. Agreement, over- and underreporting of student reports relative to parent reports about books and digital devices in the home Source OECD (2022[9]), PISA Field-Trial datasets for 2020/2021 (60 countries/economies).
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Bar chart, Measurement diagnostic.
**Physical source pages:** 33.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig13-Agreement,%20over-%20and%20underreporting%20of%20student%20reports%20relative%20to%20parent%20reports%20about%20books%20and%20digital%20devices%20in%20the%20home%20Source%20OECD%20(2022%5B9%5D),%20PISA....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=825>) · Original paper and literature note are not hosted.
### P250 - Fig. 14
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 14. Student-reported digital devices, by parent report Source OECD (2022[9]), PISA Field-Trial datasets for 2020/2021 (60 countries/economies).
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Bar chart, Measurement diagnostic.
**Physical source pages:** 34.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig14-Student-reported%20digital%20devices,%20by%20parent%20report%20Source%20OECD%20(2022%5B9%5D),%20PISA%20Field-Trial%20datasets%20for%202020%202021%20(60%20countries%20economies).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=826>) · Original paper and literature note are not hosted.
### P250 - Fig. 15
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 15. Student-reported number of parents or guardians, by parent report Conditional distribution of the number of parents or guardians Note: Agreeing student and parent reports in dark shade; discrepant reports in light shade Source: OECD (2022[9]), PISA Field-Trial datasets for 2020/2021 (14 countries/economies).
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Bar chart, Measurement diagnostic.
**Physical source pages:** 36.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig15-Student-reported%20number%20of%20parents%20or%20guardians,%20by%20parent%20report%20Conditional%20distribution%20of%20the%20number%20of%20parents%20or%20guardians%20Note%20Agreeing%20student%20and%20parent....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=827>) · Original paper and literature note are not hosted.
### P250 - Fig. 16
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 16. Highest education level (HISCED) and occupational status (HISEI) among parents/guardians Source: OECD (2022[9]), PISA Field-Trial datasets for 2020/2021 (60 countries/economies).
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Bar chart, Measurement diagnostic.
**Physical source pages:** 37.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig16-Highest%20education%20level%20(HISCED)%20and%20occupational%20status%20(HISEI)%20among%20parents%20guardians%20Source%20OECD%20(2022%5B9%5D),%20PISA%20Field-Trial%20datasets%20for%202020%202021%20(60....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=828>) · Original paper and literature note are not hosted.
### P250 - Fig. 17
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 17. Subjective and objective measures of socio-economic status Note: Each figure reports the arithmetic average, across all countries and economies with valid data, of the mean “objective” measure of socio-economic status for each value on the subjective socio-economic ladder (1-10). Source: PISA 2022 Dataset, available at OECD (2024[7]), PISA Data and Methodology, https://www.oecd.org/en/about/programmes/pisa/pisa-data.html.
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 40.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig17-Subjective%20and%20objective%20measures%20of%20socio-economic%20status%20Note%20Each%20figure%20reports%20the%20arithmetic%20average,%20across%20all%20countries%20and%20economies%20with%20valid%20data,%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=829>) · Original paper and literature note are not hosted.
### P250 - Fig. 18
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 18. Food insecurity, by subjective socio-economic status and decile of the household possession index Note: Each chart shows the arithmetic average, across all countries and economies with valid data, of the proportion of students who reported not eating “about once a week” “because there was not enough money to buy food”. The left chart reports mean proportions for each point on the subjective socio-economic ladder (1-10); the right chart for each decile of the HOMEPOS index. Source: PISA 2022 Dataset, available at OECD (2024[7]), PISA Data and Methodology, https://www.oecd.org/en/about/programmes/pisa/pisa-data.html.
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 41.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig18-Food%20insecurity,%20by%20subjective%20socio-economic%20status%20and%20decile%20of%20the%20household%20possession%20index%20Note%20Each%20chart%20shows%20the%20arithmetic%20average,%20across%20all....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=830>) · Original paper and literature note are not hosted.
### P250 - Fig. 19
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 19. Student reports on housing, by quarter of occupational status (HISEI) and number of siblings in the household Note: Numerals (1-4) refer to quarters of the HISEI index. “None”, “One”, ... refer to the number of siblings in the household, based on student reports. For the purpose of computing the average number of rooms with a bath or shower/with a flush toilet, the highest response category (3 or more”) was treated as “3”. Source: PISA 2022 Dataset.
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 42.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig19-Student%20reports%20on%20housing,%20by%20quarter%20of%20occupational%20status%20(HISEI)%20and%20number%20of%20siblings%20in%20the%20household%20Note%20Numerals%20(1-4)%20refer%20to%20quarters%20of%20the%20HISEI....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=831>) · Original paper and literature note are not hosted.
### P250 - Fig. 20
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 20. Percentage of students in the bottom quarter of the household possession index, by school disadvantage category Note: Categories of school disadvantage are based on school reports on the share of students “from disadvantaged homes”. The horizontal line indicates 25%, the average share of students from the bottom quarter of household possessions across all schools. Source: PISA 2022 Dataset.
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 44.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig20-Percentage%20of%20students%20in%20the%20bottom%20quarter%20of%20the%20household%20possession%20index,%20by%20school%20disadvantage%20category%20Note%20Categories%20of%20school%20disadvantage%20are%20based....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=832>) · Original paper and literature note are not hosted.
### P250 - Fig. 21
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Figure 21. Indicators of household possession in 2022, by school disadvantage category Selected indicators only Note: The vertical axis indicates the percentage of students who reported possessing each item; the scale differs across items. The horizontal axis indicates the share of students “from disadvantaged homes”, according to principals. Source: PISA 2022 Dataset.
**Purpose:** Measurement quality, respondent agreement, instrument comparability or SES scale validation in the paper. **Visual forms:** Composite, Measurement diagnostic.
**Physical source pages:** 45.
[PNG](<04_Measurement_and_Data_Quality/P250-Fig21-Indicators%20of%20household%20possession%20in%202022,%20by%20school%20disadvantage%20category%20Selected%20indicators%20only%20Note%20The%20vertical%20axis%20indicates%20the%20percentage%20of%20students....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=833>) · Original paper and literature note are not hosted.
### P250 - App Table A.1
**Paper:** The Measurement of Socio-economic Status in PISA: A Review of Recent Experiments and Changes
**Original caption:** Table A.1. Changes to household-possession items
**Purpose:** Detailed cross-cycle item correspondence with explicit linking and wording-change notes; structured measurement evidence. **Visual forms:** Complex table, Instrument comparison.
**Physical source pages:** 51, 52.
[PNG](<04_Measurement_and_Data_Quality/P250-AppA-TableA.1-Changes%20to%20household-possession%20items.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=834>) · Original paper and literature note are not hosted.
### P251 - Fig. 6
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 6. Study 1: Evaluation of the Brunswik Symmetry principle with regard to reasoning. Change of correlations when specific reasoning abilities (low level) were considered instead of general reasoning (medium level). Rea = Reasoning. Blue lines indicate substantially different correlations based on an equivalence testing approach. The facets of Neuroticism (-N) were reversed to be in line with the dimension of Emotional Stability.
**Purpose:** Evaluates how measurement aggregation changes correlations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 13.
[PNG](<04_Measurement_and_Data_Quality/P251-Fig06-Study%201%20Evaluation%20of%20the%20Brunswik%20Symmetry%20principle%20with%20regard%20to%20reasoning.%20Change%20of%20correlations%20when%20specific%20reasoning%20abilities%20(low%20level)%20were....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=835>) · Original paper and literature note are not hosted.
### P251 - Fig. 7
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 7. Study 1: Evaluation of the Brunswik Symmetry principle with regard to personality. Change of correlations when Stability and Plasticity (Big Two) were considered instead of GFP (both high level). Blue lines indicate substantially different correlations based on an equivalence testing approach.
**Purpose:** Evaluates how measurement aggregation changes correlations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 14.
[PNG](<04_Measurement_and_Data_Quality/P251-Fig07-Study%201%20Evaluation%20of%20the%20Brunswik%20Symmetry%20principle%20with%20regard%20to%20personality.%20Change%20of%20correlations%20when%20Stability%20and%20Plasticity%20(Big%20Two)%20were%20considered....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=836>) · Original paper and literature note are not hosted.
### P251 - Fig. 8
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 8. Study 1: Evaluation of the Brunswik Symmetry principle with regard to personality. Change of correlations when FFM dimensions (medium level) were considered instead of Stability and Plasticity (Big Two; high level). Blue lines indicate substantially different correlations based on an equivalence testing approach.
**Purpose:** Evaluates how measurement aggregation changes correlations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 14.
[PNG](<04_Measurement_and_Data_Quality/P251-Fig08-Study%201%20Evaluation%20of%20the%20Brunswik%20Symmetry%20principle%20with%20regard%20to%20personality.%20Change%20of%20correlations%20when%20FFM%20dimensions%20(medium%20level)%20were%20considered....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=837>) · Original paper and literature note are not hosted.
### P251 - Fig. 9
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 9. Study 1: Evaluation of the Brunswik Symmetry principle with regard to personality. Change of correlations when facets (low level) were considered instead of FFM dimensions (medium level). Rea = reasoning, ver = verbal, num = numerical, fig = figural. Blue lines indicate substantially different correlations based on an equivalence testing approach.
**Purpose:** Evaluates how measurement aggregation changes correlations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 15.
[PNG](<04_Measurement_and_Data_Quality/P251-Fig09-Study%201%20Evaluation%20of%20the%20Brunswik%20Symmetry%20principle%20with%20regard%20to%20personality.%20Change%20of%20correlations%20when%20facets%20(low%20level)%20were%20considered%20instead%20of%20FFM....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=838>) · Original paper and literature note are not hosted.
### P251 - Fig. 13
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 13. Study 2: Evaluation of the Brunswik Symmetry principle with regard to intelligence. Differences in correlations between the g-factor (high level) and specific abilities (medium level). Blue lines indicate substantially different correlations based on an equivalence testing approach.
**Purpose:** Evaluates how measurement aggregation changes correlations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 23.
[PNG](<04_Measurement_and_Data_Quality/P251-Fig13-Study%202%20Evaluation%20of%20the%20Brunswik%20Symmetry%20principle%20with%20regard%20to%20intelligence.%20Differences%20in%20correlations%20between%20the%20g-factor%20(high%20level)%20and%20specific....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=839>) · Original paper and literature note are not hosted.
### P251 - Fig. 14
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 14. Study 2: Evaluation of the Brunswik Symmetry principle with regard to intelligence. Differences in correlations between specific abilities (medium level) and low-level abilities. Blue lines indicate substantially different correlations based on an equivalence testing approach.
**Purpose:** Evaluates how measurement aggregation changes correlations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 24.
[PNG](<04_Measurement_and_Data_Quality/P251-Fig14-Study%202%20Evaluation%20of%20the%20Brunswik%20Symmetry%20principle%20with%20regard%20to%20intelligence.%20Differences%20in%20correlations%20between%20specific%20abilities%20(medium%20level)%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=840>) · Original paper and literature note are not hosted.
### P251 - Fig. 15
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 15. Study 2: Evaluation of the Brunswik Symmetry principle with regard to personality. Change of correlations when Stability and Plasticity (Big Two) were considered instead of GFP (both high level). Blue lines indicate substantially different correlations based on an equivalence testing approach.
**Purpose:** Evaluates how measurement aggregation changes correlations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 24.
[PNG](<04_Measurement_and_Data_Quality/P251-Fig15-Study%202%20Evaluation%20of%20the%20Brunswik%20Symmetry%20principle%20with%20regard%20to%20personality.%20Change%20of%20correlations%20when%20Stability%20and%20Plasticity%20(Big%20Two)%20were%20considered....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=841>) · Original paper and literature note are not hosted.
### P251 - Fig. 16
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 16. Study 2: Evaluation of the Brunswik Symmetry principle with regard to personality. Change of correlations when FFM dimensions (medium level) were considered instead of the Big Two (high level). Rea = reasoning, Flu = fluency, Spe = perceptual speed, Mem = memory, Ver = verbal (intelligence), Num = numerical (intelligence), Fig = figural (intelligence). Blue lines indicate substantially different correlations based on an equivalence testing approach.
**Purpose:** Evaluates how measurement aggregation changes correlations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 25.
[PNG](<04_Measurement_and_Data_Quality/P251-Fig16-Study%202%20Evaluation%20of%20the%20Brunswik%20Symmetry%20principle%20with%20regard%20to%20personality.%20Change%20of%20correlations%20when%20FFM%20dimensions%20(medium%20level)%20were%20considered....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=842>) · Original paper and literature note are not hosted.
### P252 - Fig. 2
**Paper:** The reliability paradox: Why robust cognitive tasks do not produce reliable individual differences
**Original caption:** Fig. 2 Reliability of key measures from Studies 1 and 2 combined (Total N=99–104). Red marker indicates mean group performance from sessions 1 and 2. Error bars show ± 1 standard error of measurement (SEM). The SEM is the square root of the error variance term calculated from the intraclass correlation, and can be interpreted as the 68% confidence interval for an individual’s data point. A large SEM relative to the between-subject variance contributes to poor reliability
**Purpose:** Retest reliability and SEM for cognitive task effects. **Visual forms:** Scatter, Composite.
**Physical source pages:** 7.
[PNG](<04_Measurement_and_Data_Quality/P252-Fig02-Reliability%20of%20key%20measures%20from%20Studies%201%20and%202%20combined%20(Total%20N=99-104).%20Red%20marker%20indicates%20mean%20group%20performance%20from%20sessions%201%20and%202.%20Error%20bars%20show%20%C2%B1%201....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=843>) · Original paper and literature note are not hosted.
### P252 - Fig. 3
**Paper:** The reliability paradox: Why robust cognitive tasks do not produce reliable individual differences
**Original caption:** Fig. 3 Reliability of key measures from Study 3 (N=40). Red marker indicates mean group performance from sessions 1 and 2. Error bars show ± 1 standard error of measurement. RT reaction time, SNARC Spatial-Numerical Association of Response Code
**Purpose:** Retest reliability for the additional tasks. **Visual forms:** Scatter, Composite.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P252-Fig03-Reliability%20of%20key%20measures%20from%20Study%203%20(N=40).%20Red%20marker%20indicates%20mean%20group%20performance%20from%20sessions%201%20and%202.%20Error%20bars%20show%20%C2%B1%201%20standard%20error%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=844>) · Original paper and literature note are not hosted.
### P252 - Fig. 4
**Paper:** The reliability paradox: Why robust cognitive tasks do not produce reliable individual differences
**Original caption:** Fig. 4 Relative size of variance components for reaction time (RT) measures in Studies 1 and 2 (A: Total N=99–104) and Study 3 (B: N=40). The size of the bar is normalized for the total amount of variance in the measure (see Supplementary Material E), and subdivided into variance accounted for by differences between participants (white), variance accounted for by differences between sessions (e.g., practice effects, gray), and error variance (black). The intraclass correlation (ICC) reflects the proportion of the total variance attributed to variance between individuals, and is printed above each bar. SSD Stop-Signal Delay, SSRT Stop-Signal Reaction Time, SNARC Spatial-Numerical Association of Response Code
**Purpose:** Variance decomposition explains why robust effects can be unreliable person measures. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 11.
[PNG](<04_Measurement_and_Data_Quality/P252-Fig04-Relative%20size%20of%20variance%20components%20for%20reaction%20time%20(RT)%20measures%20in%20Studies%201%20and%202%20(A%20Total%20N=99-104)%20and%20Study%203%20(B%20N=40).%20The%20size%20of%20the%20bar%20is%20normalized....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=845>) · Original paper and literature note are not hosted.
### P253 - Fig. 1
**Paper:** Validity and fairness of the PISA 2018 Global Competence assessment: an argument-based evaluation via explanatory item response models
**Original caption:** Fig. 1 Meta-analytic forest plot of testlet fixed effects.
**Purpose:** Testlet-specific item-difficulty effects assess construct-irrelevant design variation in global-competence measurement. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 8.
[PNG](<04_Measurement_and_Data_Quality/P253-Fig01-Meta-analytic%20forest%20plot%20of%20testlet%20fixed%20effects.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=846>) · Original paper and literature note are not hosted.
### P253 - App Fig. S1
**Paper:** Validity and fairness of the PISA 2018 Global Competence assessment: an argument-based evaluation via explanatory item response models
**Original caption:** Figure S1. Scree plot and parallel analysis for a representative booklet group
**Purpose:** Parallel-analysis diagnostic evaluates the unidimensionality assumption for a representative booklet group. **Visual forms:** Scree plot.
**Physical source pages:** 6.
[PNG](<04_Measurement_and_Data_Quality/P253-App-FigS1-Scree%20plot%20and%20parallel%20analysis%20for%20a%20representative%20booklet%20group.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=847>) · Original paper and literature note are not hosted.
### P253 - App Fig. S3
**Paper:** Validity and fairness of the PISA 2018 Global Competence assessment: an argument-based evaluation via explanatory item response models
**Original caption:** Figure S3. Meta-analytic forest plot grid for student-level covariate effects
**Purpose:** Booklet-specific covariate associations and pooled effects provide construct-validity evidence. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 19.
[PNG](<04_Measurement_and_Data_Quality/P253-App-FigS3-Meta-analytic%20forest%20plot%20grid%20for%20student-level%20covariate%20effects.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=848>) · Original paper and literature note are not hosted.
### P253 - App Fig. S7
**Paper:** Validity and fairness of the PISA 2018 Global Competence assessment: an argument-based evaluation via explanatory item response models
**Original caption:** Figure S7. Forest plot grid for item-level DIF effects
**Purpose:** Item-level gender DIF estimates and pooled effects assess measurement fairness. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 23.
[PNG](<04_Measurement_and_Data_Quality/P253-App-FigS7-Forest%20plot%20grid%20for%20item-level%20DIF%20effects.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=849>) · Original paper and literature note are not hosted.
### P253 - App Fig. S8
**Paper:** Validity and fairness of the PISA 2018 Global Competence assessment: an argument-based evaluation via explanatory item response models
**Original caption:** Figure S8. Caterpillar plot of item-level DIF effects (all items)
**Purpose:** Complete item-level DIF profile evaluates measurement fairness. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 24.
[PNG](<04_Measurement_and_Data_Quality/P253-App-FigS8-Caterpillar%20plot%20of%20item-level%20DIF%20effects%20(all%20items).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=850>) · Original paper and literature note are not hosted.
## 05 Methods and Algorithms
### P065 - Fig. 1
**Paper:** The impact of CT-integrated mathematics instruction on secondary school students' computational thinking skills
**Original caption:** Fig. 1. CT-Integrated Mathematics Instruction Framework.
**Purpose:** Designed CT-integrated instructional pathways explain the intervention method. **Visual forms:** Flowchart, Composite.
**Physical source pages:** 3.
[PNG](<05_Methods_and_Algorithms/P065-Fig01-CT-Integrated%20Mathematics%20Instruction%20Framework.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=260>) · Original paper and literature note are not hosted.
### P068 - Fig. 1
**Paper:** Understanding students’ mathematical achievement through feature selection and machine learning: Evidence from PISA 2018
**Original caption:** Figure 1. Overview of the analytical workflow applied to the PISA 2018 Turkish sample.
**Purpose:** Analytical pipeline linking preprocessing, feature selection and classifiers. **Visual forms:** Flowchart.
**Physical source pages:** 9.
[PNG](<05_Methods_and_Algorithms/P068-Fig01-Overview%20of%20the%20analytical%20workflow%20applied%20to%20the%20PISA%202018%20Turkish%20sample.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=261>) · Original paper and literature note are not hosted.
### P068 - Table 2
**Paper:** Understanding students’ mathematical achievement through feature selection and machine learning: Evidence from PISA 2018
**Original caption:** Chart 2 Representative studies employing ML algorithms on PISA data: A focus on mathematics achievement
**Purpose:** Structured comparison of prior PISA machine-learning studies, research aims and algorithms. **Visual forms:** Complex table.
**Physical source pages:** 5.
[PNG](<05_Methods_and_Algorithms/P068-Table02-Representative%20studies%20employing%20ML%20algorithms%20on%20PISA%20data%20A%20focus%20on%20mathematics%20achievement.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=262>) · Original paper and literature note are not hosted.
### P069 - Fig. 2
**Paper:** What students learn depends on what we teach: Curriculum content coverage and student achievement in mathematics and science
**Original caption:** Fig. 2. Mediation model: Total effect.
**Purpose:** Explains the total-effect regression decomposition used in estimation. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<05_Methods_and_Algorithms/P069-Fig02-Mediation%20model%20Total%20effect.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=263>) · Original paper and literature note are not hosted.
### P069 - Fig. 3
**Paper:** What students learn depends on what we teach: Curriculum content coverage and student achievement in mathematics and science
**Original caption:** Fig. 3. Mediation model: Direct and indirect effects.
**Purpose:** Specifies the direct and indirect regression paths estimated by the mediation method. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<05_Methods_and_Algorithms/P069-Fig03-Mediation%20model%20Direct%20and%20indirect%20effects.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=264>) · Original paper and literature note are not hosted.
### P073 - Fig. 2
**Paper:** Investigating the factors influencing secondary school students’ creative thinking based on the opportunity–propensity framework: Evidence from high-performing countries and regions
**Original caption:** Fig. 2. Research procedure.
**Purpose:** Preprocessing, hierarchical modeling and random-forest analytic stages. **Visual forms:** Flowchart.
**Physical source pages:** 7.
[PNG](<05_Methods_and_Algorithms/P073-Fig02-Research%20procedure.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=265>) · Original paper and literature note are not hosted.
### P074 - Fig. 5
**Paper:** Predicting Key Factors Influencing Students’ Creative Thinking: A Machine Learning Analysis Based on PISA 2022 Data from East Asia
**Original caption:** Fig. 5. Regression coefficient path diagram.
**Purpose:** LASSO regularization coefficient path explains feature selection. **Visual forms:** Line/profile plot.
**Physical source pages:** 9.
[PNG](<05_Methods_and_Algorithms/P074-Fig05-Regression%20coefficient%20path%20diagram.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=266>) · Original paper and literature note are not hosted.
### P082 - App Table B1
**Paper:** C017
**Original caption:** Table B1. Coding Sheet for General Article Information
**Purpose:** Structured review coding schema maps variables to operational definitions and hierarchical code lists; continuation is part of the same table. **Visual forms:** Complex table.
**Physical source pages:** 5, 6.
[PNG](<05_Methods_and_Algorithms/P082-App-TableB1-Coding%20Sheet%20for%20General%20Article%20Information.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=851>) · Original paper and literature note are not hosted.
### P082 - App Table B2
**Paper:** C017
**Original caption:** Table B2. Coding Sheet for the Relationship between Factors
**Purpose:** Structured coding schema distinguishes outcome/exposure, direction, moderation and moderator roles. **Visual forms:** Complex table.
**Physical source pages:** 7.
[PNG](<05_Methods_and_Algorithms/P082-App-TableB2-Coding%20Sheet%20for%20the%20Relationship%20between%20Factors.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=852>) · Original paper and literature note are not hosted.
### P083 - Fig. 2
**Paper:** Profiles of academic performance and perceived digital distraction in secondary classrooms
**Original caption:** Fig. 2. The three-phases analytical framework for identifying and characterizing Student profiles.
**Purpose:** Explains preprocessing, clustering, classification and interpretation pipeline. **Visual forms:** Flowchart.
**Physical source pages:** 7.
[PNG](<05_Methods_and_Algorithms/P083-Fig02-The%20three-phases%20analytical%20framework%20for%20identifying%20and%20characterizing%20Student%20profiles.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=267>) · Original paper and literature note are not hosted.
### P093 - Fig. 1
**Paper:** Using machine learning algorithms to predict students’ general self-efficacy in PISA 2018
**Original caption:** Fig. 1. Workflow of Data Preprocessing and Analysis.
**Purpose:** Explains preprocessing, model evaluation and interpretation workflow. **Visual forms:** Flowchart.
**Physical source pages:** 5.
[PNG](<05_Methods_and_Algorithms/P093-Fig01-Workflow%20of%20Data%20Preprocessing%20and%20Analysis.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=268>) · Original paper and literature note are not hosted.
### P093 - Fig. 2
**Paper:** Using machine learning algorithms to predict students’ general self-efficacy in PISA 2018
**Original caption:** Fig. 2. The Nested Cross-Validation Algorithm.
**Purpose:** Explains nested cross-validation design used by the predictive method. **Visual forms:** Flowchart.
**Physical source pages:** 8.
[PNG](<05_Methods_and_Algorithms/P093-Fig02-The%20Nested%20Cross-Validation%20Algorithm.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=269>) · Original paper and literature note are not hosted.
### P094 - Table 3
**Paper:** What teachers can do to enhance students’ well-being: Discussion
**Original caption:** Table 3. Study methods. Note: Information regarding results do not integrate all results. They have been summarized only to notify whether there were associations between student reports and student well-being, between teacher reports and student well-being, or between parents’/principals’ reports and student well-being.
**Purpose:** Structured cross-study comparison of the special issue evidence, with multiple conceptual partitions. **Visual forms:** Complex table, Structured evidence comparison.
**Physical source pages:** 4.
[PNG](<05_Methods_and_Algorithms/P094-Table03-Study%20methods.%20Note%20Information%20regarding%20results%20do%20not%20integrate%20all%20results.%20They%20have%20been%20summarized%20only%20to%20notify%20whether%20there%20were%20associations%20between....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=270>) · Original paper and literature note are not hosted.
### P095 - Table 10
**Paper:** International differences in educational equity: An assessment using the Benefit of the Doubt model
**Original caption:** Table 10 Studies of performance of education systems using BoD composite indicators at the country level including equity indicators.
**Purpose:** Comparison of prior equity-index studies and their methodological choices. **Visual forms:** Complex table.
**Physical source pages:** 14.
[PNG](<05_Methods_and_Algorithms/P095-Table10-Studies%20of%20performance%20of%20education%20systems%20using%20BoD%20composite%20indicators%20at%20the%20country%20level%20including%20equity%20indicators.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=271>) · Original paper and literature note are not hosted.
### P096 - Fig. 1
**Paper:** Pathways to equity: A mediation analysis of gender, SES, and mathematics achievement using PISA 2022 UK data
**Original caption:** Fig. 1. The total effect of student individual characteristics on mathematics achievement.
**Purpose:** Explains the total-effect path before the mediation decomposition. **Visual forms:** Path diagram.
**Physical source pages:** 5.
[PNG](<05_Methods_and_Algorithms/P096-Fig01-The%20total%20effect%20of%20student%20individual%20characteristics%20on%20mathematics%20achievement.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=272>) · Original paper and literature note are not hosted.
### P101 - Fig. 1
**Paper:** Building multicultural teaching self-efficacy: Key factors of secondary teachers’ confidence in diverse classrooms
**Original caption:** Fig. 1. Coefficient paths across lambda values.
**Purpose:** Regularization paths explain how the LASSO selection method changes coefficients with penalty strength. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<05_Methods_and_Algorithms/P101-Fig01-Coefficient%20paths%20across%20lambda%20values.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=273>) · Original paper and literature note are not hosted.
### P115 - Fig. 1
**Paper:** Is well begun half done? A latent cross-lagged study of first-year teachers’ occupational commitment, vocational satisfaction, and teacher self-efficacy
**Original caption:** Fig. 1. Competing Models. Notes. VS₁, TSE₁, and OC₁ are vocational satisfaction, teacher self-efficacy, and occupational commitment at T1; VS₂, TSE₂, and OC₂ are these variables at T2; the dotted arrows refer to the autoregressive effects; the solid arrows refer to the causal effects.
**Purpose:** Four competing longitudinal specifications explain the cross-lagged model comparison. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 6.
[PNG](<05_Methods_and_Algorithms/P115-Fig01-Competing%20Models.%20Notes.%20VS1,%20TSE1,%20and%20OC1%20are%20vocational%20satisfaction,%20teacher%20self-efficacy,%20and%20occupational%20commitment%20at%20T1;%20VS2,%20TSE2,%20and%20OC2%20are%20these....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=274>) · Original paper and literature note are not hosted.
### P116 - Fig. 1
**Paper:** School leadership and teacher stress: Unpacking the mediating role of job satisfaction and the employment status as a moderator
**Original caption:** Fig. 1. study design with the structural model.
**Purpose:** Specifies the structural and indicator model used in the analysis. **Visual forms:** Path diagram.
**Physical source pages:** 5.
[PNG](<05_Methods_and_Algorithms/P116-Fig01-study%20design%20with%20the%20structural%20model.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=275>) · Original paper and literature note are not hosted.
### P128 - Table 4
**Paper:** “Together we are better”: Professional learning networks for teachers
**Original caption:** Table 4 Research question 1 codebook.
**Purpose:** Qualitative coding framework with definition and evidence exemplar for each code. **Visual forms:** Complex table.
**Physical source pages:** 6.
[PNG](<05_Methods_and_Algorithms/P128-Table04-Research%20question%201%20codebook.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=276>) · Original paper and literature note are not hosted.
### P128 - Table 5
**Paper:** “Together we are better”: Professional learning networks for teachers
**Original caption:** Table 5 Research question 2 codebook.
**Purpose:** Multilevel category–code–definition–evidence qualitative coding framework. **Visual forms:** Complex table.
**Physical source pages:** 7.
[PNG](<05_Methods_and_Algorithms/P128-Table05-Research%20question%202%20codebook.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=277>) · Original paper and literature note are not hosted.
### P128 - Table 6
**Paper:** “Together we are better”: Professional learning networks for teachers
**Original caption:** Table 6 Research question 3 codebook.
**Purpose:** Grouped coding framework connects outcome categories with definitions and participant evidence. **Visual forms:** Complex table.
**Physical source pages:** 7.
[PNG](<05_Methods_and_Algorithms/P128-Table06-Research%20question%203%20codebook.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=278>) · Original paper and literature note are not hosted.
### P129 - Fig. 1
**Paper:** A bibliometric and thematic analysis of PISA literature for school improvement and effectiveness over two decades
**Original caption:** Fig. 1. A four-dimensional conceptual model for bibliometric review.
**Purpose:** Framework organizing the bibliometric review dimensions. **Visual forms:** Composite.
**Physical source pages:** 3.
[PNG](<05_Methods_and_Algorithms/P129-Fig01-A%20four-dimensional%20conceptual%20model%20for%20bibliometric%20review.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=279>) · Original paper and literature note are not hosted.
### P131 - Fig. 1
**Paper:** Educational equity measurement: A comparative analysis of equity indicators
**Original caption:** Fig. 1. The standard error of measurement functions for the conditions of (1) large SEM, (2) small SEM, (3) easy test and (4) difficult test.
**Purpose:** Simulation design defining four test-measurement error functions. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 7.
[PNG](<05_Methods_and_Algorithms/P131-Fig01-The%20standard%20error%20of%20measurement%20functions%20for%20the%20conditions%20of%20(1)%20large%20SEM,%20(2)%20small%20SEM,%20(3)%20easy%20test%20and%20(4)%20difficult%20test.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=280>) · Original paper and literature note are not hosted.
### P134 - Fig. 4
**Paper:** Separating prior knowledge from acquired knowledge: An individual differences analysis of PISA - learning in the digital world
**Original caption:** Fig. 4. Model Specifications for Estimating and Explaining the Residual.
**Purpose:** Model specification explains separation of prior knowledge and the residual. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<05_Methods_and_Algorithms/P134-Fig04-Model%20Specifications%20for%20Estimating%20and%20Explaining%20the%20Residual.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=281>) · Original paper and literature note are not hosted.
### P143 - Table 1
**Paper:** Adapting Fairness Interventions to Missing Values
**Original caption:** Table 1: Distribution of the synthetic data. N denotes the number of data points sampled from the group distribution.
**Purpose:** Simulation design specifies observed/missing strata with distinct multivariate/univariate generation distributions. **Visual forms:** Complex table.
**Physical source pages:** 18.
[PNG](<05_Methods_and_Algorithms/P143-Table01-Distribution%20of%20the%20synthetic%20data.%20N%20denotes%20the%20number%20of%20data%20points%20sampled%20from%20the%20group%20distribution.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=282>) · Original paper and literature note are not hosted.
### P143 - Fig. 4
**Paper:** Adapting Fairness Interventions to Missing Values
**Original caption:** Figure 4: Illustration of the three adaptive algorithms for fair linear classifiers (left) and FairMissBag (right). Adding missing indicators and affinely adaptive classification encode missing value patterns via additional input features, while missing pattern clustering partitions the dataset based on missing patterns. FairMissBag is a three-step algorithm consisting of fair resampling, using a fairness intervention to train a fair classifier on each resampled dataset, and obtaining a prediction from the ensemble of fair classifiers.
**Purpose:** Explains three adaptive algorithms and the FairMissBag ensemble workflow. **Visual forms:** Flowchart, Composite.
**Physical source pages:** 17.
[PNG](<05_Methods_and_Algorithms/P143-Fig04-Illustration%20of%20the%20three%20adaptive%20algorithms%20for%20fair%20linear%20classifiers%20(left)%20and%20FairMissBag%20(right).%20Adding%20missing%20indicators%20and%20affinely%20adaptive%20classification....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=283>) · Original paper and literature note are not hosted.
### P143 - Fig. 5
**Paper:** Adapting Fairness Interventions to Missing Values
**Original caption:** Figure 5: Plot of the synthetic data. The points on the x-axis represent data with X₂ missing.
**Purpose:** Synthetic-data construction illustrates known class, group and missing-value structure. **Visual forms:** Scatter.
**Physical source pages:** 18.
[PNG](<05_Methods_and_Algorithms/P143-Fig05-Plot%20of%20the%20synthetic%20data.%20The%20points%20on%20the%20x-axis%20represent%20data%20with%20X2%20missing.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=284>) · Original paper and literature note are not hosted.
### P147 - Fig. 2
**Paper:** A Review of Generalizability and Transportability
**Original caption:** Figure 2. Overview framework for assessing and addressing external validity bias after data collection.
**Purpose:** Methodological workflow for identifying and addressing external-validity bias. **Visual forms:** Flowchart.
**Physical source pages:** 4.
[PNG](<05_Methods_and_Algorithms/P147-Fig02-Overview%20framework%20for%20assessing%20and%20addressing%20external%20validity%20bias%20after%20data%20collection.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=285>) · Original paper and literature note are not hosted.
### P151 - Fig. 1
**Paper:** Decomposition Methods in Economics
**Original caption:** Figure 1: Relationship Between Proportions and Quantiles
**Purpose:** Explains how estimated counterfactual proportions map to distributional quantiles. **Visual forms:** Conceptual curve.
**Physical source pages:** 110.
[PNG](<05_Methods_and_Algorithms/P151-Fig01-Relationship%20Between%20Proportions%20and%20Quantiles.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=286>) · Original paper and literature note are not hosted.
### P151 - Table 1
**Paper:** Decomposition Methods in Economics
**Original caption:** Table 1. Maintained Assumptions and Properties of Major Decomposition Methodologies. Note: ᵃ Unless otherwise indicated, the different methodologies appeal to a simple counterfactual treatment. ᵇ In some applications, the counterfactual is an average over time periods or over countries.
**Purpose:** Structured comparison of decomposition families, identification assumptions, properties and applications. **Visual forms:** Complex table.
**Physical source pages:** 113.
[PNG](<05_Methods_and_Algorithms/P151-Table01-Maintained%20Assumptions%20and%20Properties%20of%20Major%20Decomposition%20Methodologies.%20Note%20a%20Unless%20otherwise%20indicated,%20the%20different%20methodologies%20appeal%20to%20a%20simple....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=287>) · Original paper and literature note are not hosted.
### P151 - Fig. 2
**Paper:** Decomposition Methods in Economics
**Original caption:** Figure 2: Inverting Globally
**Purpose:** Explains global inversion of counterfactual distributions. **Visual forms:** Conceptual curve.
**Physical source pages:** 111.
[PNG](<05_Methods_and_Algorithms/P151-Fig02-Inverting%20Globally.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=288>) · Original paper and literature note are not hosted.
### P151 - Fig. 3
**Paper:** Decomposition Methods in Economics
**Original caption:** Figure 3: RIF Regressions: Inverting Locally
**Purpose:** Explains the local inversion underlying RIF-regression decomposition. **Visual forms:** Conceptual curve.
**Physical source pages:** 112.
[PNG](<05_Methods_and_Algorithms/P151-Fig03-RIF%20Regressions%20Inverting%20Locally.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=289>) · Original paper and literature note are not hosted.
### P152 - Fig. 1
**Paper:** Domain Adaptation under Missingness Shift
**Original caption:** Figure 1: DAMS with UCAR. The source and target data are drawn from the same P(X, Y ), but differ in how ξ (and hence Xe) takes its value. Shaded nodes are observed. Observed covariates are generated as Xe = X ⊙ ξ. The undirected edge between X and Y indicates that they can have an arbitrary bidirectional relationship.
**Purpose:** Explains the domain-adaptation missingness-shift setup. **Visual forms:** Causal/graphical model.
**Physical source pages:** 3.
[PNG](<05_Methods_and_Algorithms/P152-Fig01-DAMS%20with%20UCAR.%20The%20source%20and%20target%20data%20are%20drawn%20from%20the%20same%20P(X,%20Y%20),%20but%20differ%20in%20how%20%CE%BE%20(and%20hence%20Xe)%20takes%20its%20value.%20Shaded%20nodes%20are%20observed.%20Observed....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=290>) · Original paper and literature note are not hosted.
### P154 - Fig. 1
**Paper:** EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE
**Original caption:** Figure 1: Illustration of Partial VAE encoder architecture.
**Purpose:** Partial variational-autoencoder architectures for arbitrary observed subsets. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 4.
[PNG](<05_Methods_and_Algorithms/P154-Fig01-Illustration%20of%20Partial%20VAE%20encoder%20architecture.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=291>) · Original paper and literature note are not hosted.
### P157 - Fig. 1
**Paper:** Estimating Social and Ethnic Inequality in School Surveys: Biases from Child Misreporting and Parent Nonresponse
**Original caption:** Figure 1. Illustration of linear regression with classical ME in x-axis
**Purpose:** Explains attenuation produced by classical covariate measurement error. **Visual forms:** Scatter.
**Physical source pages:** 3.
[PNG](<05_Methods_and_Algorithms/P157-Fig01-Illustration%20of%20linear%20regression%20with%20classical%20ME%20in%20x-axis.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=292>) · Original paper and literature note are not hosted.
### P157 - Fig. 2
**Paper:** Estimating Social and Ethnic Inequality in School Surveys: Biases from Child Misreporting and Parent Nonresponse
**Original caption:** Figure 2. Illustration of linear regression with selection on y-axis
**Purpose:** Explains bias produced by selection on the outcome. **Visual forms:** Scatter.
**Physical source pages:** 4.
[PNG](<05_Methods_and_Algorithms/P157-Fig02-Illustration%20of%20linear%20regression%20with%20selection%20on%20y-axis.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=293>) · Original paper and literature note are not hosted.
### P158 - Fig. 1
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 1 Flow chart for an overview of the chapters in this article including connections between chapters for the three research questions: RQ 1 (green), RQ 2 (purple), RQ 3 (red)
**Purpose:** Maps the evidence-synthesis workflow and research-question connections. **Visual forms:** Flowchart.
**Physical source pages:** 3.
[PNG](<05_Methods_and_Algorithms/P158-Fig01-Flow%20chart%20for%20an%20overview%20of%20the%20chapters%20in%20this%20article%20including%20connections%20between%20chapters%20for%20the%20three%20research%20questions%20RQ%201%20(green),%20RQ%202%20(purple),%20RQ%203%20(red).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=294>) · Original paper and literature note are not hosted.
### P158 - Table 8
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Table 8 Overview of existing reviews and meta analyses
**Purpose:** Multi-part comparison of review scope, selection, effects, moderators and quality methods. **Visual forms:** Complex table, Structured evidence comparison.
**Physical source pages:** 95, 96, 97.
[PNG](<05_Methods_and_Algorithms/P158-Table08-Overview%20of%20existing%20reviews%20and%20meta%20analyses.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=295>) · Original paper and literature note are not hosted.
### P158 - Table 9
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Table 9 Comparison of existing reviews and meta analyses to this work
**Purpose:** Structured comparison establishes differences in synthesis methods and scope. **Visual forms:** Complex table, Structured evidence comparison.
**Physical source pages:** 98.
[PNG](<05_Methods_and_Algorithms/P158-Table09-Comparison%20of%20existing%20reviews%20and%20meta%20analyses%20to%20this%20work.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=296>) · Original paper and literature note are not hosted.
### P159 - Table B3
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Table B3: Included Studies in Literature Comparison. Notes: This table lists the studies included in our literature comparison. N is the total sample across all arms included in our comparison. All studies compare AI access to a no-AI control on unassisted assessments.
**Purpose:** Structured comparison of prior AI-access experiments by study design, domain, sample, and outcome. **Visual forms:** Complex table.
**Physical source pages:** 71.
[PNG](<05_Methods_and_Algorithms/P159-TableB3-Included%20Studies%20in%20Literature%20Comparison.%20Notes%20This%20table%20lists%20the%20studies%20included%20in%20our%20literature%20comparison.%20N%20is%20the%20total%20sample%20across%20all%20arms%20included%20in....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=297>) · Original paper and literature note are not hosted.
### P161 - Fig. 2
**Paper:** Extramural ICT factors impact adolescents’ academic performance and well-being differently: Types of self-regulated learners also matter
**Original caption:** Fig. 2 Illustrative process of data analysis
**Purpose:** Analysis workflow covers data quality, latent-profile selection and multigroup regression. **Visual forms:** Flowchart.
**Physical source pages:** 17.
[PNG](<05_Methods_and_Algorithms/P161-Fig02-Illustrative%20process%20of%20data%20analysis.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=298>) · Original paper and literature note are not hosted.
### P166 - Fig. 1
**Paper:** Imputation and missing indicators for handling missing data in the development and deployment of clinical prediction models: A simulation study
**Original caption:** Figure 1. Directed acyclic graphs for four missingness structures.
**Purpose:** Explains simulated missingness-generating structures. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 4.
[PNG](<05_Methods_and_Algorithms/P166-Fig01-Directed%20acyclic%20graphs%20for%20four%20missingness%20structures.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=299>) · Original paper and literature note are not hosted.
### P166 - Table 1
**Paper:** Imputation and missing indicators for handling missing data in the development and deployment of clinical prediction models: A simulation study
**Original caption:** Table 1. Imputation and validation strategies.
**Purpose:** Compares imputation/validation strategies and deployment data requirements. **Visual forms:** Complex table.
**Physical source pages:** 6.
[PNG](<05_Methods_and_Algorithms/P166-Table01-Imputation%20and%20validation%20strategies.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=300>) · Original paper and literature note are not hosted.
### P166 - Fig. 2
**Paper:** Imputation and missing indicators for handling missing data in the development and deployment of clinical prediction models: A simulation study
**Original caption:** Figure 2. Simulation procedure, step-by-step.
**Purpose:** Stepwise simulation generation and analysis algorithm. **Visual forms:** Flowchart.
**Physical source pages:** 5.
[PNG](<05_Methods_and_Algorithms/P166-Fig02-Simulation%20procedure,%20step-by-step.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=301>) · Original paper and literature note are not hosted.
### P167 - Fig. 3
**Paper:** Imputation Strategies Under Clinical Presence: Impact on Algorithmic Fairness
**Original caption:** Figure 3: Graphical summary of clinical missingness enforcement in the simulation experiments. Note that our simulations’ choices result in missingness in the marginalised
**Purpose:** Summarizes how the simulation scenarios induce clinical missingness. **Visual forms:** Flowchart.
**Physical source pages:** 7.
[PNG](<05_Methods_and_Algorithms/P167-Fig03-Graphical%20summary%20of%20clinical%20missingness%20enforcement%20in%20the%20simulation%20experiments.%20Note%20that%20our%20simulations%E2%80%99%20choices%20result%20in%20missingness%20in%20the%20marginalised.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=302>) · Original paper and literature note are not hosted.
### P170 - Fig. 3
**Paper:** Introduction: Administrative Burden as a Mechanism of Inequality in Policy Implementation
**Original caption:** Figure 3. The Medicare Maze Source: Hanoch and Rice 2006. Reprinted with permission.
**Purpose:** Medicare eligibility and enrollment decision workflow illustrates procedural complexity. **Visual forms:** Flowchart.
**Physical source pages:** 18.
[PNG](<05_Methods_and_Algorithms/P170-Fig03-The%20Medicare%20Maze%20Source%20Hanoch%20and%20Rice%202006.%20Reprinted%20with%20permission.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=303>) · Original paper and literature note are not hosted.
### P173 - Table 1
**Paper:** jomo: A Flexible Package for Two-level Joint Modelling Multiple Imputation
**Original caption:** Table 1: Summary of subfunctions used by the main "umbrella function" jomo, given the type of imputation model, the level at which missing data occur, and the type of the variables with missing data.
**Purpose:** Function-selection matrix for imputation model levels and variable types. **Visual forms:** Complex table.
**Physical source pages:** 3.
[PNG](<05_Methods_and_Algorithms/P173-Table01-Summary%20of%20subfunctions%20used%20by%20the%20main%20umbrella%20function%20jomo,%20given%20the%20type%20of%20imputation%20model,%20the%20level%20at%20which%20missing%20data%20occur,%20and%20the%20type%20of%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=304>) · Original paper and literature note are not hosted.
### P174 - Fig. 2
**Paper:** Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach
**Original caption:** FIGURE 2.—An illustration of the estimation of the effect of unions for male high school graduates with 10 to 30 years of experience.
**Purpose:** Illustrates reweighting for the union contribution before formal method derivation. **Visual forms:** Density, Composite.
**Physical source pages:** 11.
[PNG](<05_Methods_and_Algorithms/P174-Fig02-An%20illustration%20of%20the%20estimation%20of%20the%20effect%20of%20unions%20for%20male%20high%20school%20graduates%20with%2010%20to%2030%20years%20of%20experience.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=305>) · Original paper and literature note are not hosted.
### P174 - Fig. 3
**Paper:** Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach
**Original caption:** FIGURE 3.-An illustration of the estimation of the effect of the minimum wage for female high school dropouts with 20 years or less of experience.
**Purpose:** Illustrates the minimum-wage splice used in density decomposition. **Visual forms:** Density, Composite.
**Physical source pages:** 18.
[PNG](<05_Methods_and_Algorithms/P174-Fig03-An%20illustration%20of%20the%20estimation%20of%20the%20effect%20of%20the%20minimum%20wage%20for%20female%20high%20school%20dropouts%20with%2020%20years%20or%20less%20of%20experience.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=306>) · Original paper and literature note are not hosted.
### P174 - Table II
**Paper:** Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach
**Original caption:** TABLE II WEIGHTS USED IN DENSITY DECOMPOSITION. Note: The weighting functions ψ, ψ_{u|x}, ψ_x, ψ′, ψ_{x|u} and ψ_u are defined by equations (19), (8), (11), (25), (27), and (26), respectively.
**Purpose:** Formula-valued weighting matrix explains forward and reverse sequential density decomposition. **Visual forms:** Complex table.
**Physical source pages:** 23.
[PNG](<05_Methods_and_Algorithms/P174-TableII-WEIGHTS%20USED%20IN%20DENSITY%20DECOMPOSITION.%20Note%20The%20weighting%20functions%20%CF%88,%20%CF%88_%7Bu%20x%7D,%20%CF%88_x,%20%CF%88%E2%80%B2,%20%CF%88_%7Bx%20u%7D%20and%20%CF%88_u%20are%20defined%20by%20equations%20(19),%20(8),%20(11),%20(25),%20(27),%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=307>) · Original paper and literature note are not hosted.
### P175 - Fig. Extended Data 1
**Paper:** Machine learning and phone data can improve targeting of humanitarian aid
**Original caption:** Extended Data Fig. 1 | Overview of targeting methodology. a) Regional targeting. Satellite imagery of Togo59 is used to construct micro-estimates of poverty (middle)16, which are overlayed with population data to produce canton-level estimates of wealth. Individuals registered in the 100 poorest cantons (right) are eligible for benefits. Inset images © 2019 Google. b) Individual targeting. A machine-learning algorithm is trained using representative survey data to predict consumption from features of phone use (Methods, ‘ Machine-learning methods’). The algorithm constructs poverty scores that are correlated with ground-truth measures of consumption (left). Subscribers who register for the program in targeted cantons with estimated consumption less than USD $1.25/day are eligible for benefits (right). The red distribution shows the predicted wealth distribution of the entire population of Togo; the blue distribution shows the predicted wealth distribution in the 100 poorest cantons; and the green section indicates the predicted wealth distribution of Novissi beneficiaries.
**Purpose:** Explains regional satellite mapping and individual phone-data targeting. **Visual forms:** Map, Flowchart, Scatter, Density, Composite.
**Physical source pages:** 18.
[PNG](<05_Methods_and_Algorithms/P175-FigExtended%20Data%201-Extended%20Data%20Fig.%201%20Overview%20of%20targeting%20methodology.%20a)%20Regional%20targeting.%20Satellite%20imagery%20of%20Togo59%20is%20used%20to%20construct%20micro-estimates%20of%20poverty....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=308>) · Original paper and literature note are not hosted.
### P176 - Fig. 1
**Paper:** mirt: A Multidimensional Item Response Theory Package for the R Environment
**Original caption:** Figure 1: Unidimensional and two-factor test information plots for the LSAT7 data.
**Purpose:** Demonstrates unidimensional and multidimensional test-information visualization produced by mirt. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 27.
[PNG](<05_Methods_and_Algorithms/P176-Fig01-Unidimensional%20and%20two-factor%20test%20information%20plots%20for%20the%20LSAT7%20data.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=309>) · Original paper and literature note are not hosted.
### P176 - Fig. 2
**Paper:** mirt: A Multidimensional Item Response Theory Package for the R Environment
**Original caption:** Figure 2: Combined probability plot for each LSAT7 item.
**Purpose:** Demonstrates mirt item-response probability output on the LSAT7 example. **Visual forms:** Line/profile plot.
**Physical source pages:** 28.
[PNG](<05_Methods_and_Algorithms/P176-Fig02-Combined%20probability%20plot%20for%20each%20LSAT7%20item.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=310>) · Original paper and literature note are not hosted.
### P177 - Table 1
**Paper:** Missing data and prediction: the pattern submodel
**Original caption:** Table 1. Comparing imputation methods for an out-of-sample individual with missing data
**Purpose:** Compares required inputs, advantages and drawbacks of competing prediction-time missingness methods. **Visual forms:** Complex table.
**Physical source pages:** 3.
[PNG](<05_Methods_and_Algorithms/P177-Table01-Comparing%20imputation%20methods%20for%20an%20out-of-sample%20individual%20with%20missing%20data.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=311>) · Original paper and literature note are not hosted.
### P177 - Table 2
**Paper:** Missing data and prediction: the pattern submodel
**Original caption:** Table 2. Comparison of pattern submodels and complete case submodels
**Purpose:** Compares pattern-specific mathematical models and their complete-case counterparts. **Visual forms:** Complex table.
**Physical source pages:** 6.
[PNG](<05_Methods_and_Algorithms/P177-Table02-Comparison%20of%20pattern%20submodels%20and%20complete%20case%20submodels.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=312>) · Original paper and literature note are not hosted.
### P180 - Table 1
**Paper:** On the Treatment of Missing Data in Background Questionnaires in Educational Large-Scale Assessments: An Evaluation of Different Procedures
**Original caption:** TABLE 1. Schematic Illustration of the MIM
**Purpose:** Schematic recoding map explains missing indicators and extra-category representations. **Visual forms:** Complex table.
**Physical source pages:** 6.
[PNG](<05_Methods_and_Algorithms/P180-Table01-Schematic%20Illustration%20of%20the%20MIM.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=313>) · Original paper and literature note are not hosted.
### P180 - Table 2
**Paper:** On the Treatment of Missing Data in Background Questionnaires in Educational Large-Scale Assessments: An Evaluation of Different Procedures
**Original caption:** TABLE 2. Expected Values for the Estimated Variances and Covariances Under the MIM Organized by the Pattern of Missing Data and the Method Used to Treat Missing Data Under MCAR. Note. MIMᵣ₌₀ = MIM with estimates based on only the complete cases; MIMᵣ₌₁ = MIM with estimates based on only the incomplete cases (i.e., with missing data x₁); MIM-PE = MIM with pairwise estimation; MIM-LD = MIM with listwise deletion; MIM-MI = MIM with nested MI; p = proportion of missing data in x₁; γ₁, γ₂ = parameters of the imputation model for x₁; V, C = bias in variances and covariances (see Online Supplement A).
**Purpose:** Analytical expected-moment formulas compare missingness patterns and procedures. **Visual forms:** Complex table.
**Physical source pages:** 10.
[PNG](<05_Methods_and_Algorithms/P180-Table02-Expected%20Values%20for%20the%20Estimated%20Variances%20and%20Covariances%20Under%20the%20MIM%20Organized%20by%20the%20Pattern%20of%20Missing%20Data%20and%20the%20Method%20Used%20to%20Treat%20Missing%20Data%20Under....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=314>) · Original paper and literature note are not hosted.
### P180 - Table 3
**Paper:** On the Treatment of Missing Data in Background Questionnaires in Educational Large-Scale Assessments: An Evaluation of Different Procedures
**Original caption:** TABLE 3. Simulated Conditions in Studies 1 and 2. ᵃ The strength of the missing data mechanism denotes the amount of variance explained in the latent response propensity of x₂ when the data are MAR.
**Purpose:** Simulation-factor design contrasts two method-evaluation studies. **Visual forms:** Complex table.
**Physical source pages:** 14.
[PNG](<05_Methods_and_Algorithms/P180-Table03-Simulated%20Conditions%20in%20Studies%201%20and%202.%20a%20The%20strength%20of%20the%20missing%20data%20mechanism%20denotes%20the%20amount%20of%20variance%20explained%20in%20the%20latent%20response%20propensity%20of%20x2....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=315>) · Original paper and literature note are not hosted.
### P180 - Table 4
**Paper:** On the Treatment of Missing Data in Background Questionnaires in Educational Large-Scale Assessments: An Evaluation of Different Procedures
**Original caption:** TABLE 4. Scaling and Imputation Procedures in Study 1. Note. M = number of imputations (nested imputations are denoted by “×”).
**Purpose:** Method comparison specifies scaling, conditional imputation models and nested imputation counts. **Visual forms:** Complex table.
**Physical source pages:** 15.
[PNG](<05_Methods_and_Algorithms/P180-Table04-Scaling%20and%20Imputation%20Procedures%20in%20Study%201.%20Note.%20M%20=%20number%20of%20imputations%20(nested%20imputations%20are%20denoted%20by%20%E2%80%9C%C3%97%E2%80%9D).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=316>) · Original paper and literature note are not hosted.
### P189 - Fig. 1
**Paper:** Specification curve analysis
**Original caption:** Fig. 1 | Sets of possible specifications as perceived by researchers. a, The set of specifications reported in an article are a small subset of those the researcher would consider valid to report. b, Different researchers may have similar views on the set of valid specifications but report quite different subsets of them. c, Different researchers may also disagree on the set of specifications they consider valid.
**Purpose:** Specification-set diagrams explain selection of valid nonredundant model alternatives. **Visual forms:** Conceptual diagram, Composite.
**Physical source pages:** 2.
[PNG](<05_Methods_and_Algorithms/P189-Fig01-Sets%20of%20possible%20specifications%20as%20perceived%20by%20researchers.%20a,%20The%20set%20of%20specifications%20reported%20in%20an%20article%20are%20a%20small%20subset%20of%20those%20the%20researcher%20would....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=317>) · Original paper and literature note are not hosted.
### P189 - Table 1
**Paper:** Specification curve analysis
**Original caption:** Table 1 | Original and alternative reasonable specifications used to test whether hurricanes with more feminine names were associated with more deaths
**Purpose:** Structured decision-by-decision comparison defines the model specification universe. **Visual forms:** Complex table.
**Physical source pages:** 4.
[PNG](<05_Methods_and_Algorithms/P189-Table01-Original%20and%20alternative%20reasonable%20specifications%20used%20to%20test%20whether%20hurricanes%20with%20more%20feminine%20names%20were%20associated%20with%20more%20deaths.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=318>) · Original paper and literature note are not hosted.
### P197 - Fig. U01
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** No original caption supplied.
**Purpose:** Explains separate random-number streams for parallel simulation. **Visual forms:** Conceptual diagram.
**Physical source pages:** 10.
[PNG](<05_Methods_and_Algorithms/P197-FigU01-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=319>) · Original paper and literature note are not hosted.
### P197 - Table 1
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** TABLE 1 Key steps and decisions in the planning, coding, analysis and reporting of simulation studies
**Purpose:** Distinct multi-section simulation planning, execution, analysis and reporting checklist. **Visual forms:** Complex table, Structured checklist.
**Physical source pages:** 3.
[PNG](<05_Methods_and_Algorithms/P197-Table01-Key%20steps%20and%20decisions%20in%20the%20planning,%20coding,%20analysis%20and%20reporting%20of%20simulation%20studies.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=320>) · Original paper and literature note are not hosted.
### P197 - Fig. 1
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** FIGURE 1 The impacts of bias and empirical SE on root MSE and coverage of nominal 95% confidence intervals, compared for three methods: Method A is unbiased but imprecise; Method B is biased (independent of nobs) and more precise; Method C is biased (with bias ∝ √1∕nobs) and the same precision as method B. The comparison of root MSE and coverage depends on the choice of nobs; the constant bias of method B dominates its increasingly poor MSE and coverage as nobs increases [Colour figure can be viewed at wileyonlinelibrary.com]
**Purpose:** Illustrates how bias and precision jointly determine performance metrics. **Visual forms:** Composite, Line plot.
**Physical source pages:** 14.
[PNG](<05_Methods_and_Algorithms/P197-Fig01-The%20impacts%20of%20bias%20and%20empirical%20SE%20on%20root%20MSE%20and%20coverage%20of%20nominal%2095%25%20confidence%20intervals,%20compared%20for%20three%20methods%20Method%20A%20is%20unbiased%20but%20imprecise;%20Method....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=321>) · Original paper and literature note are not hosted.
### P197 - Fig. 2
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** FIGURE 2 Visualisation of the true hazard rate over follow-up time in the two data-generating mechanisms. Black (flat) lines are for the first data-generating mechanism, where γ = 1; Red curves are for the second, where γ = 1.5 [Colour figure can be viewed at wileyonlinelibrary.com]
**Purpose:** Explains the simulation data-generating hazard functions. **Visual forms:** Line plot.
**Physical source pages:** 18.
[PNG](<05_Methods_and_Algorithms/P197-Fig02-Visualisation%20of%20the%20true%20hazard%20rate%20over%20follow-up%20time%20in%20the%20two%20data-generating%20mechanisms.%20Black%20(flat)%20lines%20are%20for%20the%20first%20data-generating%20mechanism,%20where%20%CE%B3....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=322>) · Original paper and literature note are not hosted.
### P197 - Table 3
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** TABLE 3 Possible targets of a simulation study and relevant performance measures
**Purpose:** Maps statistical tasks to simulation targets, performance measures and example analyses. **Visual forms:** Complex table, Method comparison.
**Physical source pages:** 7.
[PNG](<05_Methods_and_Algorithms/P197-Table03-Possible%20targets%20of%20a%20simulation%20study%20and%20relevant%20performance%20measures.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=323>) · Original paper and literature note are not hosted.
### P197 - Table 5
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** TABLE 5 The different datasets that may be involved in a simulation study
**Purpose:** Distinguishes conceptual and generated datasets in the simulation workflow. **Visual forms:** Complex table.
**Physical source pages:** 9.
[PNG](<05_Methods_and_Algorithms/P197-Table05-The%20different%20datasets%20that%20may%20be%20involved%20in%20a%20simulation%20study.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=324>) · Original paper and literature note are not hosted.
### P197 - Table 6
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** TABLE 6 Performance measures: definitions, estimates and Monte Carlo standard errors
**Purpose:** Multi-part comparison of performance definitions, estimators and Monte Carlo standard errors. **Visual forms:** Complex table, Formula comparison.
**Physical source pages:** 13.
[PNG](<05_Methods_and_Algorithms/P197-Table06-Performance%20measures%20definitions,%20estimates%20and%20Monte%20Carlo%20standard%20errors.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=325>) · Original paper and literature note are not hosted.
### P199 - Table 5
**Paper:** When Do Covariates Matter? And Which Ones, and How Much?
**Original caption:** Table 5: Understanding the Decomposition of δrace Note: Estimates of Γrace educ and Γrace AFQT are computed as the coefficient on a black indicator from least-squares estimation of equations with education/AFQT as the dependent variable and a constant and Hispanic and age controls included. Estimates of β2,educ and β2,AFQT are taken from the full specification of the wage equation, which includes all X1 variables (a constant, black and Hispanic indicators, and age) and both X2 variables (AFQT and years of schooling). All equations include the same sample used above and include 1,749 observations.
**Purpose:** Explains conditional versus sequential decomposition components through a two-panel mathematical comparison. **Visual forms:** Complex table.
**Physical source pages:** 40.
[PNG](<05_Methods_and_Algorithms/P199-Table05-Understanding%20the%20Decomposition%20of%20%CE%B4race%20Note%20Estimates%20of%20%CE%93race%20educ%20and%20%CE%93race%20AFQT%20are%20computed%20as%20the%20coefficient%20on%20a%20black%20indicator%20from%20least-squares....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=326>) · Original paper and literature note are not hosted.
### P201 - Fig. 2
**Paper:** A psychometrics of individual differences in experimental tasks
**Original caption:** Fig. 2 Analysis of a single task. a Sample effects from one block of data for all individuals with 95% CIs. b Model estimates from one block of data for all individuals with 95% credible intervals. c Hierarchical shrinkage in practice. The bottom row shows sample estimates from one block; the middle row shows the same from the hierarchical model; the top row shows sample estimates from all the blocks. The shrinkage from the hierarchical model attenuates the variability so that it matches that from much larger samples
**Purpose:** Explains hierarchical shrinkage and its effect on person estimates. **Visual forms:** Composite.
**Physical source pages:** 6.
[PNG](<05_Methods_and_Algorithms/P201-Fig02-Analysis%20of%20a%20single%20task.%20a%20Sample%20effects%20from%20one%20block%20of%20data%20for%20all%20individuals%20with%2095%25%20CIs.%20b%20Model%20estimates%20from%20one%20block%20of%20data%20for%20all%20individuals%20with....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=853>) · Original paper and literature note are not hosted.
### P201 - Fig. 5
**Paper:** A psychometrics of individual differences in experimental tasks
**Original caption:** Fig. 5 Prior specifications. a The critical priors are on σ²θ, σ²ω, and σ²γ. Shown are the density functions for the standard deviations (σθ, σω, σγ). b The implied prior for the correlation coefficient ρ
**Purpose:** Specifies prior distributions used in the Bayesian model comparison. **Visual forms:** Density/histogram/boxplot, Composite.
**Physical source pages:** 14.
[PNG](<05_Methods_and_Algorithms/P201-Fig05-Prior%20specifications.%20a%20The%20critical%20priors%20are%20on%20%CF%832%CE%B8,%20%CF%832%CF%89,%20and%20%CF%832%CE%B3.%20Shown%20are%20the%20density%20functions%20for%20the%20standard%20deviations%20(%CF%83%CE%B8,%20%CF%83%CF%89,%20%CF%83%CE%B3).%20b%20The%20implied%20prior%20for....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=854>) · Original paper and literature note are not hosted.
### P204 - Table 1
**Paper:** Balancing Covariates via Propensity Score Weighting
**Original caption:** Table 1: Examples of balancing weights and corresponding target population and estimand under different h.
**Purpose:** Structured mathematical comparison of target populations, tilting functions, estimands and paired weight formulas; formulas are within a method-comparison table. **Visual forms:** Complex table.
**Physical source pages:** 8.
[PNG](<05_Methods_and_Algorithms/P204-Table01-Examples%20of%20balancing%20weights%20and%20corresponding%20target%20population%20and%20estimand%20under%20different%20h.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=855>) · Original paper and literature note are not hosted.
### P204 - Fig. 1
**Paper:** Balancing Covariates via Propensity Score Weighting
**Original caption:** Figure 1: Overlap weights for two normally-distributed groups with different means. In the upper panel, the left and right solid lines, the thin and thick dashed lines represent the density of the covariate in the control, treated, combined (h(x) = 1), and overlap weighted (h(x) = e(x)(1 − e(x))) populations, respectively. In the lower panel, the two solid lines represent w0(x), w1(x) and the dashed line represents h(x) = e(x)(1 − e(x)).
**Purpose:** Method illustration explains the overlap target and weighting functions for hypothetical normal groups. **Visual forms:** Density plot, Composite.
**Physical source pages:** 12.
[PNG](<05_Methods_and_Algorithms/P204-Fig01-Overlap%20weights%20for%20two%20normally-distributed%20groups%20with%20different%20means.%20In%20the%20upper%20panel,%20the%20left%20and%20right%20solid%20lines,%20the%20thin%20and%20thick%20dashed%20lines%20represent....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=856>) · Original paper and literature note are not hosted.
### P206 - Table 3
**Paper:** Beyond traditional interviews: Psychometric analysis of asynchronous video interviews for personality and interview performance evaluation using machine learning
**Original caption:** Table 3 Summary of the 79 audio and 57 visual extracted handcrafted features. Note. We did not extract handcrafted verbal features (see Bags of Words and S-BERT embeddings in main text, instead).
**Purpose:** Modality-by-feature taxonomy documents the model inputs. **Visual forms:** Complex table.
**Physical source pages:** 8.
[PNG](<05_Methods_and_Algorithms/P206-Table03-Summary%20of%20the%2079%20audio%20and%2057%20visual%20extracted%20handcrafted%20features.%20Note.%20We%20did%20not%20extract%20handcrafted%20verbal%20features%20(see%20Bags%20of%20Words%20and%20S-BERT%20embeddings%20in....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=857>) · Original paper and literature note are not hosted.
### P211 - Table 1
**Paper:** Comparisons of statistical methods for handling attrition in a follow-up visit with complex survey sampling
**Original caption:** TABLE 1 Statistical software and corresponding procedures for design-based analysis of complex sample survey data for linear and generalized linear models with complete data
**Purpose:** Distinctive hierarchical comparison of model types, software, procedures and implementation workflows; retain complete merged row groups. **Visual forms:** Complex table.
**Physical source pages:** 6.
[PNG](<05_Methods_and_Algorithms/P211-Table01-Statistical%20software%20and%20corresponding%20procedures%20for%20design-based%20analysis%20of%20complex%20sample%20survey%20data%20for%20linear%20and%20generalized%20linear%20models%20with%20complete%20data.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=858>) · Original paper and literature note are not hosted.
### P211 - Table 2
**Paper:** Comparisons of statistical methods for handling attrition in a follow-up visit with complex survey sampling
**Original caption:** TABLE 2 Statistical software and corresponding procedures for design-based analysis of sample survey data in the presence of attrition
**Purpose:** Structured comparison of weighting, imputation and likelihood implementations across software with nested workflow descriptions. **Visual forms:** Complex table.
**Physical source pages:** 7.
[PNG](<05_Methods_and_Algorithms/P211-Table02-Statistical%20software%20and%20corresponding%20procedures%20for%20design-based%20analysis%20of%20sample%20survey%20data%20in%20the%20presence%20of%20attrition.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=859>) · Original paper and literature note are not hosted.
### P212 - Fig. 1
**Paper:** Computer-based personality judgments are more accurate than those made by humans
**Original caption:** Fig. 1. Methodology used to obtain computer-based judgments and estimate the self-other agreement. Participants and their Likes are represented as a matrix, where entries are set to 1 if there exists an association between a participant and a Like and 0 otherwise (second panel). The matrix is used to fit five LASSO linear regression models (16), one for each self-rated Big Five personality trait (third panel). A 10-fold cross-validation is applied to avoid overfitting: the sample is randomly divided into 10 equal-sized subsets; 9 subsets are used to train the model (step 1), which is then applied to the remaining subset to predict the personality score (step 2). This procedure is repeated 10 times to predict personality for the entire sample. The models are built on participants having at least 20 Likes. To estimate the accuracy achievable with less than 20 Likes, we applied the regression models to random subsets of 1–19 Likes for all participants.
**Purpose:** Prediction workflow. **Visual forms:** Flowchart, Composite.
**Physical source pages:** 2.
[PNG](<05_Methods_and_Algorithms/P212-Fig01-Methodology%20used%20to%20obtain%20computer-based%20judgments%20and%20estimate%20the%20self-other%20agreement.%20Participants%20and%20their%20Likes%20are%20represented%20as%20a%20matrix,%20where%20entries%20are....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=860>) · Original paper and literature note are not hosted.
### P213 - App Table A1.1
**Paper:** COVID-19 school closures, learning losses and intergenerational mobility
**Original caption:** Table A1.1. Parameters for simulations by income level. Note: Values represent country-level averages. Notes: (*) Global Economic Prospects January 11th 2022 update (https://www.worldbank.org/en/publication/macro-poverty-outlook), with the regional average imputed if no country value was available for 2020 or 2021. The table provides an overview of parameters by country income level, though simulation parameters are applied at a country level. The World Bank’s Harmonized Learning Outcome (HLO) puts learning data from international and regional assessments on a comparable scale. The data can be accessed here. We assume the learning gains will vary from 20 to 50 learning points depending on the country’s income level, as explained in Azevedo, et al. (2021).
**Purpose:** Hierarchically partitioned parameter design links learning productivity, closure exposure, mitigation, explicit loss formulas and macroeconomic inputs across three scenario blocks. **Visual forms:** Complex table, Scenario comparison.
**Physical source pages:** 3.
[PNG](<05_Methods_and_Algorithms/P213-App-TableA1.1-Parameters%20for%20simulations%20by%20income%20level.%20Note%20Values%20represent%20country-level%20averages.%20Notes%20(%201%20asterisk%20)%20Global%20Economic%20Prospects%20January%2011th%202022....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=861>) · Original paper and literature note are not hosted.
### P213 - App Fig. A2.1
**Paper:** COVID-19 school closures, learning losses and intergenerational mobility
**Original caption:** Figure A2.1: Diagram Representing the Composition of Multicohorts, in the Period Measurement and in the Synthetic Cohort. Source: Author’s illustration.
**Purpose:** Explains construction of a synthetic cohort from period-specific observations, a methodological diagram. **Visual forms:** Matrix, Cohort diagram.
**Physical source pages:** 7.
[PNG](<05_Methods_and_Algorithms/P213-App-FigA2.1-Diagram%20Representing%20the%20Composition%20of%20Multicohorts,%20in%20the%20Period%20Measurement%20and%20in%20the%20Synthetic%20Cohort.%20Source%20Author%E2%80%99s%20illustration.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=862>) · Original paper and literature note are not hosted.
### P218 - Fig. 1
**Paper:** Estimating Measurement Quality in Digital Trace Data and Surveys Using the MultiTrait MultiMethod Model
**Original caption:** Figure 1. Visual representation of the MTMM model estimated. Circles represents latent variables with the T latent variables representing traits and the M ones representing different methods used. Observed variables are represented by squares. First loadings are fixed to 1 for identification, and correlated errors between traits are added. Residuals are not shown for ease of reading.
**Purpose:** Defines the MTMM measurement decomposition. **Visual forms:** Path diagram.
**Physical source pages:** 7.
[PNG](<05_Methods_and_Algorithms/P218-Fig01-Visual%20representation%20of%20the%20MTMM%20model%20estimated.%20Circles%20represents%20latent%20variables%20with%20the%20T%20latent%20variables%20representing%20traits%20and%20the%20M%20ones%20representing....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=863>) · Original paper and literature note are not hosted.
### P222 - Fig. 3
**Paper:** From byproduct to design factor: on validating the interpretation of process indicators based on log data
**Original caption:** Fig. 3 Process of completing an item represented as a sequence of actions and states (SERP = search engine results page). One possible alternative process, leaving the item without interaction with the stimulus. * Additional log events not considered in this example.
**Purpose:** Explains state/action reconstruction from events. **Visual forms:** Flowchart.
**Physical source pages:** 8.
[PNG](<05_Methods_and_Algorithms/P222-Fig03-Process%20of%20completing%20an%20item%20represented%20as%20a%20sequence%20of%20actions%20and%20states%20(SERP%20=%20search%20engine%20results%20page).%20One%20possible%20alternative%20process,%20leaving%20the%20item....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=864>) · Original paper and literature note are not hosted.
### P223 - Fig. 1
**Paper:** Generating group-level scores under response accuracy-time conditional dependence
**Original caption:** Fig. 1 Measurement Models: M1 (left) and M2 (right)
**Purpose:** Defines the baseline and residual-dependence measurement models. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 9.
[PNG](<05_Methods_and_Algorithms/P223-Fig01-Measurement%20Models%20M1%20(left)%20and%20M2%20(right).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=865>) · Original paper and literature note are not hosted.
### P223 - Fig. 2
**Paper:** Generating group-level scores under response accuracy-time conditional dependence
**Original caption:** Fig. 2 IRT-LRMs: M3 (left), M4 (middle), and M5 (right)
**Purpose:** Shows nested latent regression specifications used in the model comparisons. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 10.
[PNG](<05_Methods_and_Algorithms/P223-Fig02-IRT-LRMs%20M3%20(left),%20M4%20(middle),%20and%20M5%20(right).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=866>) · Original paper and literature note are not hosted.
### P228 - Fig. 2
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Fig. 2 Illustration of states between log events and indicators computed using reconstructed sequences of states
**Purpose:** Explains FSM indicator extraction. **Visual forms:** Flowchart.
**Physical source pages:** 11.
[PNG](<05_Methods_and_Algorithms/P228-Fig02-Illustration%20of%20states%20between%20log%20events%20and%20indicators%20computed%20using%20reconstructed%20sequences%20of%20states.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=867>) · Original paper and literature note are not hosted.
### P228 - Table 2
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Table 2 Description of log events provided by the assessment platform available for the extraction of item-level response times
**Purpose:** Structured taxonomy maps raw events to assessment-process meanings. **Visual forms:** Complex table.
**Physical source pages:** 18.
[PNG](<05_Methods_and_Algorithms/P228-Table02-Description%20of%20log%20events%20provided%20by%20the%20assessment%20platform%20available%20for%20the%20extraction%20of%20item-level%20response%20times.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=868>) · Original paper and literature note are not hosted.
### P228 - Table 3
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Table 3 Annotated example table with raw log data for one test taker starting with question “Q02” and augmented information from the FSM approach
**Purpose:** Worked event-to-state transformation includes raw/augmented grouped columns and extensive explanation. **Visual forms:** Complex table.
**Physical source pages:** 21, 22.
[PNG](<05_Methods_and_Algorithms/P228-Table03-Annotated%20example%20table%20with%20raw%20log%20data%20for%20one%20test%20taker%20starting%20with%20question%20%E2%80%9CQ02%E2%80%9D%20and%20augmented%20information%20from%20the%20FSM%20approach.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=869>) · Original paper and literature note are not hosted.
### P228 - Fig. 4
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Fig. 4 State diagram for a FSM differentiating states “Reading (stem) & first question”, answering “Remaining questions” and
**Purpose:** Explains FSM indicator extraction. **Visual forms:** Flowchart.
**Physical source pages:** 19.
[PNG](<05_Methods_and_Algorithms/P228-Fig04-State%20diagram%20for%20a%20FSM%20differentiating%20states%20%E2%80%9CReading%20(stem)%20&%20first%20question%E2%80%9D,%20answering%20%E2%80%9CRemaining%20questions%E2%80%9D%20and.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=870>) · Original paper and literature note are not hosted.
### P228 - Fig. 5
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Fig. 5 Schematic view of the construction of the states “Reading stem & answering Q01”, “Answering Q02”, “Answering Q03” and “Time after last answer”
**Purpose:** Explains FSM indicator extraction. **Visual forms:** Flowchart.
**Physical source pages:** 19.
[PNG](<05_Methods_and_Algorithms/P228-Fig05-Schematic%20view%20of%20the%20construction%20of%20the%20states%20%E2%80%9CReading%20stem%20&%20answering%20Q01%E2%80%9D,%20%E2%80%9CAnswering%20Q02%E2%80%9D,%20%E2%80%9CAnswering%20Q03%E2%80%9D%20and%20%E2%80%9CTime%20after%20last%20answer%E2%80%9D.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=871>) · Original paper and literature note are not hosted.
### P228 - Fig. 6
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Fig. 6 State diagram for a FSM differentiating states “Reading & Q…”, “Answering Q…” and “Time after last answer” used for the decomposition of item-level response times
**Purpose:** Explains FSM indicator extraction. **Visual forms:** Flowchart.
**Physical source pages:** 20.
[PNG](<05_Methods_and_Algorithms/P228-Fig06-State%20diagram%20for%20a%20FSM%20differentiating%20states%20%E2%80%9CReading%20&%20Q...%E2%80%9D,%20%E2%80%9CAnswering%20Q...%E2%80%9D%20and%20%E2%80%9CTime%20after%20last%20answer%E2%80%9D%20used%20for%20the%20decomposition%20of%20item-level%20response%20times.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=872>) · Original paper and literature note are not hosted.
### P237 - Fig. 21.1
**Paper:** Modeling Item Responses with Response Times as Collateral Information in Large-Scale Educational Assessments
**Original caption:** Fig. 21.1 Graphical representations of the models for responses only (2 PM), the hierarchical model with a correlation between the latent variables (HM), and the extended hierarchical model with cross-loadings from the response times to the latent ability variable (eHM)
**Purpose:** Compares alternative psychometric model structures for time as collateral information. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 12.
[PNG](<05_Methods_and_Algorithms/P237-Fig21.1-Graphical%20representations%20of%20the%20models%20for%20responses%20only%20(2%20PM),%20the%20hierarchical%20model%20with%20a%20correlation%20between%20the%20latent%20variables%20(HM),%20and%20the%20extended....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=873>) · Original paper and literature note are not hosted.
### P238 - Fig. 1
**Paper:** Profiling low-proficiency science students in the Philippines using machine learning
**Original caption:** Fig. 1 The data processing methodology is generally composed of data preparation and machine learning (ML) modeling. To find the optimal computational model, the whole data processing pipeline is performed for different sets of hyperparameters, for different machine learning approaches.
**Purpose:** Machine-learning data-preparation/training/evaluation pipeline. **Visual forms:** Flowchart.
**Physical source pages:** 4.
[PNG](<05_Methods_and_Algorithms/P238-Fig01-The%20data%20processing%20methodology%20is%20generally%20composed%20of%20data%20preparation%20and%20machine%20learning%20(ML)%20modeling.%20To%20find%20the%20optimal%20computational%20model,%20the%20whole%20data....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=874>) · Original paper and literature note are not hosted.
### P238 - Fig. 3
**Paper:** Profiling low-proficiency science students in the Philippines using machine learning
**Original caption:** Fig. 3 The designed Random Forest model for classifying poor and better-performing students given a set of 71 variables. It is composed of n = 500 decision trees called estimators with a maximum tree depth of 20. Each input to the estimator uses only a subset of variables equal to ceil(log271) or 7 variables. This minimizes the model overfitting due to the original large number of variables.
**Purpose:** Random-forest ensemble structure and voting procedure. **Visual forms:** Architecture diagram, Flowchart.
**Physical source pages:** 6.
[PNG](<05_Methods_and_Algorithms/P238-Fig03-The%20designed%20Random%20Forest%20model%20for%20classifying%20poor%20and%20better-performing%20students%20given%20a%20set%20of%2071%20variables.%20It%20is%20composed%20of%20n%20=%20500%20decision%20trees%20called....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=875>) · Original paper and literature note are not hosted.
### P241 - Fig. 1
**Paper:** Review of inverse probability weighting for dealing with missing data
**Original caption:** Figure 1. Linear regressions on Y on X. Solid line represents true population regression slope; broken line represents estimated regression slope.
**Purpose:** Six-panel method illustration of selection bias and IPW; removes unrelated following body paragraph misidentified as the caption. **Visual forms:** Scatter, Composite.
**Physical source pages:** 5.
[PNG](<05_Methods_and_Algorithms/P241-Fig01-Linear%20regressions%20on%20Y%20on%20X.%20Solid%20line%20represents%20true%20population%20regression%20slope;%20broken%20line%20represents%20estimated%20regression%20slope.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=876>) · Original paper and literature note are not hosted.
### P247 - Table 1
**Paper:** Taking PISA Seriously: How Accurate are Low-Stakes Exams?
**Original caption:** Table 1 Classification of items Note: C and I stand for correct and incorrect answers, respectively. For response content, “.” denotes that no answer is given. For time spent, “.” denotes that no time spent on these items. S: Serious, No-Res: No Response, T-L-T: Too-little-time, Non-Reach: Non-reached, M: Missing
**Purpose:** Worked schematic response/timing sequence explains the authors’ item classification rules. **Visual forms:** Complex table.
**Physical source pages:** 11.
[PNG](<05_Methods_and_Algorithms/P247-Table01-Classification%20of%20items%20Note%20C%20and%20I%20stand%20for%20correct%20and%20incorrect%20answers,%20respectively.%20For%20response%20content,%20%E2%80%9C.%E2%80%9D%20denotes%20that%20no%20answer%20is%20given.%20For%20time%20spent,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=877>) · Original paper and literature note are not hosted.
## 06 Main Findings
### P065 - Fig. 4
**Paper:** The impact of CT-integrated mathematics instruction on secondary school students' computational thinking skills
**Original caption:** Fig. 4. Pre- and Post-test Mean CT Scores by Group.
**Purpose:** Primary group comparison of CT question scores. **Visual forms:** Scatter.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P065-Fig04-Pre-%20and%20Post-test%20Mean%20CT%20Scores%20by%20Group.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=327>) · Original paper and literature note are not hosted.
### P068 - Table 1
**Paper:** Understanding students’ mathematical achievement through feature selection and machine learning: Evidence from PISA 2018
**Original caption:** Table 1 Classification performance (mean ± standard deviation) across feature selection methods, feature set sizes, and classification algorithms
**Purpose:** Main classifier comparison crosses feature-set sizes, selection procedures and metrics. **Visual forms:** Complex table.
**Physical source pages:** 16.
[PNG](<06_Main_Findings/P068-Table01-Classification%20performance%20(mean%20%C2%B1%20standard%20deviation)%20across%20feature%20selection%20methods,%20feature%20set%20sizes,%20and%20classification%20algorithms.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=328>) · Original paper and literature note are not hosted.
### P068 - Fig. 2
**Paper:** Understanding students’ mathematical achievement through feature selection and machine learning: Evidence from PISA 2018
**Original caption:** Figure 2. Top 10 variables identified by Boruta, Mutual Information, and ReliefF feature selection methods (Red: School-related variables, Black: Student-related variables).
**Purpose:** Overlap of predictors selected by three competing selection procedures. **Visual forms:** Composite.
**Physical source pages:** 14.
[PNG](<06_Main_Findings/P068-Fig02-Top%2010%20variables%20identified%20by%20Boruta,%20Mutual%20Information,%20and%20ReliefF%20feature%20selection%20methods%20(Red%20School-related%20variables,%20Black%20Student-related%20variables).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=329>) · Original paper and literature note are not hosted.
### P068 - Fig. 3
**Paper:** Understanding students’ mathematical achievement through feature selection and machine learning: Evidence from PISA 2018
**Original caption:** Figure 3. ROC curves for the classification models trained on the top 20 features selected by Mutual Information based on PV1MATH.
**Purpose:** ROC comparison of main predictive classifiers. **Visual forms:** Line/profile plot.
**Physical source pages:** 18.
[PNG](<06_Main_Findings/P068-Fig03-ROC%20curves%20for%20the%20classification%20models%20trained%20on%20the%20top%2020%20features%20selected%20by%20Mutual%20Information%20based%20on%20PV1MATH.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=330>) · Original paper and literature note are not hosted.
### P068 - Fig. 4
**Paper:** Understanding students’ mathematical achievement through feature selection and machine learning: Evidence from PISA 2018
**Original caption:** Figure 4. Conditional SHAP feature importance rankings for the Random forest model using the top 20 features selected by Mutual Information based on PV1MATH.
**Purpose:** Primary model feature-importance rankings. **Visual forms:** Bar chart.
**Physical source pages:** 19.
[PNG](<06_Main_Findings/P068-Fig04-Conditional%20SHAP%20feature%20importance%20rankings%20for%20the%20Random%20forest%20model%20using%20the%20top%2020%20features%20selected%20by%20Mutual%20Information%20based%20on%20PV1MATH.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=331>) · Original paper and literature note are not hosted.
### P069 - Fig. 4
**Paper:** What students learn depends on what we teach: Curriculum content coverage and student achievement in mathematics and science
**Original caption:** Fig. 4. Forest plot of the cycle-specific and synthesized results of the total, indirect, and direct effects.
**Purpose:** Primary cycle-specific and pooled curriculum-achievement results. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P069-Fig04-Forest%20plot%20of%20the%20cycle-specific%20and%20synthesized%20results%20of%20the%20total,%20indirect,%20and%20direct%20effects.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=332>) · Original paper and literature note are not hosted.
### P072 - Fig. 1
**Paper:** Exploring the relationship between reading literacy and creativity competence: A multilevel modeling approach
**Original caption:** Fig. 1. Results of two-level path analysis with standardized estimates. Note. * p < .001. Aggregated variables at the school level are treated as latent variables in Mplus
**Purpose:** Primary two-level standardized reading-creativity results. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P072-Fig01-Results%20of%20two-level%20path%20analysis%20with%20standardized%20estimates.%20Note.%201%20asterisk%20p%20less%20than%20.001.%20Aggregated%20variables%20at%20the%20school%20level%20are%20treated%20as%20latent%20variables%20in%20Mplus.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=333>) · Original paper and literature note are not hosted.
### P073 - Fig. 3
**Paper:** Investigating the factors influencing secondary school students’ creative thinking based on the opportunity–propensity framework: Evidence from high-performing countries and regions
**Original caption:** Fig. 3. Top 10 most important variables.
**Purpose:** Primary ranking of the ten most important predictors. **Visual forms:** Bar chart.
**Physical source pages:** 9.
[PNG](<06_Main_Findings/P073-Fig03-Top%2010%20most%20important%20variables.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=334>) · Original paper and literature note are not hosted.
### P073 - Fig. 4
**Paper:** Investigating the factors influencing secondary school students’ creative thinking based on the opportunity–propensity framework: Evidence from high-performing countries and regions
**Original caption:** Fig. 4. Importance of opportunity factors.
**Purpose:** Importance of opportunity predictors. **Visual forms:** Bar chart.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P073-Fig04-Importance%20of%20opportunity%20factors.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=335>) · Original paper and literature note are not hosted.
### P073 - Fig. 5
**Paper:** Investigating the factors influencing secondary school students’ creative thinking based on the opportunity–propensity framework: Evidence from high-performing countries and regions
**Original caption:** Fig. 5. Importance of propensity factors.
**Purpose:** Importance of propensity predictors. **Visual forms:** Bar chart.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P073-Fig05-Importance%20of%20propensity%20factors.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=336>) · Original paper and literature note are not hosted.
### P074 - Fig. 7
**Paper:** Predicting Key Factors Influencing Students’ Creative Thinking: A Machine Learning Analysis Based on PISA 2022 Data from East Asia
**Original caption:** Fig. 7. Ranking of importance of factors influencing creative thinking.
**Purpose:** Random-forest predictor importance is a primary empirical result. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P074-Fig07-Ranking%20of%20importance%20of%20factors%20influencing%20creative%20thinking.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=337>) · Original paper and literature note are not hosted.
### P077 - Fig. 1
**Paper:** Digital media use and availability can mitigate the digital divide: Meta-analytic insights from PISA 2022
**Original caption:** Fig. 1. Forest plot of the standardized beta coefficient representing the relationship between SES and self-efficacy in digital competencies (ICTEFFIC) Note. Forest plot of country-specific effect sizes with 95 % confidence intervals. Each square represents the estimated effect size for a given country, with horizontal bars indicating the corresponding 95 % confidence interval. Effect sizes are statistically significant if the confidence interval does not cross zero. The diamond at the bottom of the figure represents the overall summary effect derived from the random-effects model. Country names are represented using their three-letter ISO codes. Model estimates held constant after accounting for gender as a control variable.
**Purpose:** Country-specific estimates and pooled SES–digital-efficacy association are primary results. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P077-Fig01-Forest%20plot%20of%20the%20standardized%20beta%20coefficient%20representing%20the%20relationship%20between%20SES%20and%20self-efficacy%20in%20digital%20competencies%20(ICTEFFIC)%20Note.%20Forest%20plot%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=338>) · Original paper and literature note are not hosted.
### P077 - Fig. 2
**Paper:** Digital media use and availability can mitigate the digital divide: Meta-analytic insights from PISA 2022
**Original caption:** Fig. 2. Forest plot of the standardized beta coefficient representing the relationship between SES and practices regarding online information (ICTINFO) Note. Forest plot of country-specific effect sizes with 95 % confidence intervals. Each square represents the estimated effect size for a given country, with horizontal bars indicating the corresponding 95 % confidence interval. Effect sizes are statistically significant if the confidence interval does not cross zero. The diamond at the bottom of the figure represents the overall summary effect derived from the random-effects model. Country names are represented using their three-letter ISO codes Model estimates held constant after accounting for gender as a control variable.
**Purpose:** Country-specific estimates and pooled SES–information-practices association are primary results. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P077-Fig02-Forest%20plot%20of%20the%20standardized%20beta%20coefficient%20representing%20the%20relationship%20between%20SES%20and%20practices%20regarding%20online%20information%20(ICTINFO)%20Note.%20Forest%20plot%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=339>) · Original paper and literature note are not hosted.
### P080 - Fig. 2
**Paper:** Examining the relationship between online chatting and PISA reading literacy trends (2000-2018)
**Original caption:** Fig. 2. Trend changes/changes for South Korea and the UK.
**Purpose:** Illustrates estimated trends for two countries. **Visual forms:** Composite, Line plot.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P080-Fig02-Trend%20changes%20changes%20for%20South%20Korea%20and%20the%20UK.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=340>) · Original paper and literature note are not hosted.
### P080 - Fig. 3
**Paper:** Examining the relationship between online chatting and PISA reading literacy trends (2000-2018)
**Original caption:** Fig. 3. Correlation between trend changes in online chatting and reading literacy
Country size determines the size of each bubble; six largest countries are indicated explicitly; reading literacy trend changes are estimates from multilevel modeling
**Purpose:** Primary cross-country trend association. **Visual forms:** Bubble scatterplot.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P080-Fig03-Correlation%20between%20trend%20changes%20in%20online%20chatting%20and%20reading%20literacy%20Country%20size%20determines%20the%20size%20of%20each%20bubble;%20six%20largest%20countries%20are%20indicated%20explicitly;....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=341>) · Original paper and literature note are not hosted.
### P091 - Fig. 1
**Paper:** Subjective well-being and educational performance in PISA: A copula based approach
**Original caption:** Fig. 1. Scatter plot of observed and 1000 simulated values for PISA scores and the SWB construct from the Gaussian copula.
**Purpose:** Observed and fitted copula relationship between achievement and well-being. **Visual forms:** Scatter.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P091-Fig01-Scatter%20plot%20of%20observed%20and%201000%20simulated%20values%20for%20PISA%20scores%20and%20the%20SWB%20construct%20from%20the%20Gaussian%20copula.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=342>) · Original paper and literature note are not hosted.
### P091 - Fig. 2
**Paper:** Subjective well-being and educational performance in PISA: A copula based approach
**Original caption:** Fig. 2. Scatter plot of the cumulative distribution function of 5000 simulated values from the Gaussian copula with parameter θ = 0.005 and marginal distributions specified as gamma and normal.
**Purpose:** Simulated cumulative distribution under the selected Gaussian copula. **Visual forms:** Scatter.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P091-Fig02-Scatter%20plot%20of%20the%20cumulative%20distribution%20function%20of%205000%20simulated%20values%20from%20the%20Gaussian%20copula%20with%20parameter%20%CE%B8%20=%200.005%20and%20marginal%20distributions%20specified%20as%20gamma%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=343>) · Original paper and literature note are not hosted.
### P093 - Fig. 4
**Paper:** Using machine learning algorithms to predict students’ general self-efficacy in PISA 2018
**Original caption:** Fig. 4. Relative Importance of the Top-Ranking Predictors in the XGBoost Model.
**Purpose:** Primary XGBoost predictor-importance results. **Visual forms:** Bar chart.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P093-Fig04-Relative%20Importance%20of%20the%20Top-Ranking%20Predictors%20in%20the%20XGBoost%20Model.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=344>) · Original paper and literature note are not hosted.
### P094 - Table 5
**Paper:** What teachers can do to enhance students’ well-being: Discussion
**Original caption:** Table 5. Key findings and methodological strengths.
**Purpose:** Structured cross-study comparison of the special issue evidence, with multiple conceptual partitions. **Visual forms:** Complex table, Structured evidence comparison.
**Physical source pages:** 5.
[PNG](<06_Main_Findings/P094-Table05-Key%20findings%20and%20methodological%20strengths.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=345>) · Original paper and literature note are not hosted.
### P095 - Fig. 3
**Paper:** International differences in educational equity: An assessment using the Benefit of the Doubt model
**Original caption:** Fig. 3. ECI for OECD educational systems under the unconditional and conditional DD-BoD models.
**Purpose:** Primary geographical comparison of education-system equity. **Visual forms:** Map, Composite.
**Physical source pages:** 11.
[PNG](<06_Main_Findings/P095-Fig03-ECI%20for%20OECD%20educational%20systems%20under%20the%20unconditional%20and%20conditional%20DD-BoD%20models.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=346>) · Original paper and literature note are not hosted.
### P099 - Fig. 1
**Paper:** Understanding disparities: Student and school factors associated with U.S. students’ achievement in reading, mathematics, and science
**Original caption:** Fig. 1. Variance Decomposition Across Model Specifications. Note. Bars represent the proportion of total variance explained by fixed effects (R2) at each model stage, averaged across the 100 fits per model. Circles connected by dashed lines show the intraclass correlation coefficient (ICC), indicating the proportion of residual variance occurring between schools. ICCs decline as school-level predictors are added (Models 2–5), demonstrating that measured school characteristics account for part of the between-school variance. The slight rebound at Model 6 (Random Slopes) reflects the redistribution of variance from the random intercept into random slopes. Model specifications: Null (unconditional); Student-Level (individual predictors only); Full (adds school-level predictors); Contextual (group-mean centered individual predictors with school-mean aggregates); Cross-Level Interactions (tests moderation effects); Random Slopes (allows gender, English vs. Spanish, and within-school parental education to vary by school).
**Purpose:** Explains overall variance contributions across primary model specifications. **Visual forms:** Bar chart, Line/profile plot, Composite.
**Physical source pages:** 11.
[PNG](<06_Main_Findings/P099-Fig01-Variance%20Decomposition%20Across%20Model%20Specifications.%20Note.%20Bars%20represent%20the%20proportion%20of%20total%20variance%20explained%20by%20fixed%20effects%20(R2)%20at%20each%20model%20stage,%20averaged%20across....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=347>) · Original paper and literature note are not hosted.
### P099 - Fig. 2
**Paper:** Understanding disparities: Student and school factors associated with U.S. students’ achievement in reading, mathematics, and science
**Original caption:** Fig. 2. Student Background Characteristics and Achievement: Standardized Effects Across Domains. Note. Standardized effect sizes were computed by dividing each unstandardized coefficient from the Full Model (Model 3) by the pooled outcome standard deviation for each subject. Error bars show 95% confidence intervals based on Rubin’s rules pooling across 10 plausible values X 10 imputations. Reference categories: Gender (Female), Home Language (Spanish), Immigration Status (Native-born).
**Purpose:** Primary standardized background associations across three subjects. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 12.
[PNG](<06_Main_Findings/P099-Fig02-Student%20Background%20Characteristics%20and%20Achievement%20Standardized%20Effects%20Across%20Domains.%20Note.%20Standardized%20effect%20sizes%20were%20computed%20by%20dividing%20each%20unstandardized%20coefficient....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=348>) · Original paper and literature note are not hosted.
### P101 - Table 3
**Paper:** Building multicultural teaching self-efficacy: Key factors of secondary teachers’ confidence in diverse classrooms
**Original caption:** Table 3 Variable selection by LASSO method and post-selection significance. Note. All continuous variables were standardized before entering into LASSO models. ✓ = Variable selected (non-zero coefficient); × = Not statistically significant (p > .05) in post-selection logistic regression.
**Purpose:** Selection agreement across three LASSO approaches identifies the key retained factors. **Visual forms:** Complex table, Matrix.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P101-Table03-Variable%20selection%20by%20LASSO%20method%20and%20post-selection%20significance.%20Note.%20All%20continuous%20variables%20were%20standardized%20before%20entering%20into%20LASSO%20models.%20%E2%9C%93%20=%20Variable%20selected....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=349>) · Original paper and literature note are not hosted.
### P115 - Fig. 2
**Paper:** Is well begun half done? A latent cross-lagged study of first-year teachers’ occupational commitment, vocational satisfaction, and teacher self-efficacy
**Original caption:** Fig. 2. Cross-lagged Relationships among Vocational Satisfaction, Teacher Self-efficacy, and Occupational Commitment. Notes. For clarity’s sake, the non-significant paths, factor loadings, and path coefficients of the controlled variables are not shown in this figure; the dotted arrows refer to the autoregressive effects; the solid arrows refer to the causal effects; *p < .05, ***p < .001.
**Purpose:** Primary fitted longitudinal relationships among satisfaction, efficacy and commitment. **Visual forms:** Path diagram.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P115-Fig02-Cross-lagged%20Relationships%20among%20Vocational%20Satisfaction,%20Teacher%20Self-efficacy,%20and%20Occupational%20Commitment.%20Notes.%20For%20clarity%E2%80%99s%20sake,%20the%20non-significant%20paths,%20factor....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=350>) · Original paper and literature note are not hosted.
### P122 - Fig. 2
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 2. Teachers who report that student “Assessment methods in reading” were an area of emphasis in their initial formal education and training (%). Note: In 2021, the incidence of missing values on the indicators of interest is disproportionately high in some countries (namely, Italy, Lithuania and Germany). We still prefer to use the 2021 data, because these data are the most recent (and post-Covid) as well as because they include important indicators on student assessment which were not available in precedent waves of the survey. Also, we highlight that for the three countries of interest, the 2021 estimates do not differ significantly from the previous ones, this being an ‘indirect’ indication that missingness may have occurred “at random” with respect to the outcomes of interest. Country labels: AUS Australia, AUT Austria, BEL Belgium, BGR Bulgaria, CAB Canada (only Alberta), CAN Canada, CZE the Czech Republic (Czechia), DEU Germany, DNK Denmark, ESP Spain, EST Estonia, FIN Finland, FRA France, HRV Croatia, HUN Hungary, IRL Ireland, ITA Italy, LTU Lithuania, LVA Latvia, NLD the Netherlands, NOR Norway, PRT Portugal, ROU Romania SGP Singapore, SVK Slovakia, SVN Slovenia, SWE Sweden, UK the United Kingdom, and finally, USA the United States of America.
**Purpose:** Cross-national assessment-training and assessment-practice findings. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P122-Fig02-Teachers%20who%20report%20that%20student%20%E2%80%9CAssessment%20methods%20in%20reading%E2%80%9D%20were%20an%20area%20of%20emphasis%20in%20their%20initial%20formal%20education%20and%20training%20(%25).%20Note%20In%202021,%20the%20incidence%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=351>) · Original paper and literature note are not hosted.
### P122 - Fig. 3
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 3. Teachers who do not assign reading as part of students’ homework (%).
**Purpose:** Cross-national assessment-training and assessment-practice findings. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P122-Fig03-Teachers%20who%20do%20not%20assign%20reading%20as%20part%20of%20students%E2%80%99%20homework%20(%25).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=352>) · Original paper and literature note are not hosted.
### P122 - Fig. 4
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 4. Teachers who monitor whether or not reading homework was completed (%).
**Purpose:** Cross-national assessment-training and assessment-practice findings. **Visual forms:** Line/profile plot.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P122-Fig04-Teachers%20who%20monitor%20whether%20or%20not%20reading%20homework%20was%20completed%20(%25).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=353>) · Original paper and literature note are not hosted.
### P122 - Fig. 5
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 5. Teachers who discuss homework in class (%). Note: Authors’ elaboration on PIRLS teacher questionnaire data. There are three missing values in 2021 (Hungary, Ireland, and United States).
**Purpose:** Cross-national assessment-training and assessment-practice findings. **Visual forms:** Line/profile plot.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P122-Fig05-Teachers%20who%20discuss%20homework%20in%20class%20(%25).%20Note%20Authors%E2%80%99%20elaboration%20on%20PIRLS%20teacher%20questionnaire%20data.%20There%20are%20three%20missing%20values%20in%202021%20(Hungary,%20Ireland,%20and%20United....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=354>) · Original paper and literature note are not hosted.
### P122 - Fig. 6
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 6. Teachers trained on assessment during their ITE and their CPD (%), 2021. Note: Authors’ elaboration on PIRLS teacher questionnaire data. The horizontal and vertical dashed lines show the median values of the CPD and ITE indicators, respectively.
**Purpose:** Cross-national assessment-training and assessment-practice findings. **Visual forms:** Scatter.
**Physical source pages:** 9.
[PNG](<06_Main_Findings/P122-Fig06-Teachers%20trained%20on%20assessment%20during%20their%20ITE%20and%20their%20CPD%20(%25),%202021.%20Note%20Authors%E2%80%99%20elaboration%20on%20PIRLS%20teacher%20questionnaire%20data.%20The%20horizontal%20and%20vertical%20dashed%20lines....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=355>) · Original paper and literature note are not hosted.
### P122 - Fig. 7
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 7. Teachers trained on assessment during their ITE/CPD and a) monitoring homework or b) discussing homework in class (%), 2021. Note: Authors’ elaboration on PIRLS teacher questionnaire data. The horizontal and vertical dashed lines show the median values of the assessment practices and ITE indicators, respectively.
**Purpose:** Cross-national assessment-training and assessment-practice findings. **Visual forms:** Scatter.
**Physical source pages:** 9.
[PNG](<06_Main_Findings/P122-Fig07-Teachers%20trained%20on%20assessment%20during%20their%20ITE%20CPD%20and%20a)%20monitoring%20homework%20or%20b)%20discussing%20homework%20in%20class%20(%25),%202021.%20Note%20Authors%E2%80%99%20elaboration%20on%20PIRLS%20teacher....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=356>) · Original paper and literature note are not hosted.
### P122 - Fig. 8
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 8. Teachers engaged in CPD initiatives on assessment over the last year (%), 2021.Note: Authors’ elaboration on TALIS data.
**Purpose:** Cross-national assessment-training and assessment-practice findings. **Visual forms:** Line/profile plot.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P122-Fig08-Teachers%20engaged%20in%20CPD%20initiatives%20on%20assessment%20over%20the%20last%20year%20(%25),%202021.Note%20Authors%E2%80%99%20elaboration%20on%20TALIS%20data.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=357>) · Original paper and literature note are not hosted.
### P122 - Fig. 9
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 9. Teachers who declare a high level of need for CPD on assessment (%).Note: Authors’ elaboration on TALIS data. Many countries are not part of the survey in 2008.
**Purpose:** Cross-national assessment-training and assessment-practice findings. **Visual forms:** Line/profile plot.
**Physical source pages:** 11.
[PNG](<06_Main_Findings/P122-Fig09-Teachers%20who%20declare%20a%20high%20level%20of%20need%20for%20CPD%20on%20assessment%20(%25).Note%20Authors%E2%80%99%20elaboration%20on%20TALIS%20data.%20Many%20countries%20are%20not%20part%20of%20the%20survey%20in%202008.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=358>) · Original paper and literature note are not hosted.
### P122 - Fig. 10
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 10. Percentages of teachers who declare the use of a high variety of assessment methods.Note: Authors’ elaboration on TALIS data.
**Purpose:** Cross-national assessment-training and assessment-practice findings. **Visual forms:** Line/profile plot.
**Physical source pages:** 12.
[PNG](<06_Main_Findings/P122-Fig10-Percentages%20of%20teachers%20who%20declare%20the%20use%20of%20a%20high%20variety%20of%20assessment%20methods.Note%20Authors%E2%80%99%20elaboration%20on%20TALIS%20data.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=359>) · Original paper and literature note are not hosted.
### P122 - Fig. 11
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 11. Teachers trained on assessment during their ITE and their CPD (%), 2018. Note: Authors’ elaboration on TALIS data. The horizontal and vertical dashed lines show the median values of the CPD and ITE indicators, respectively. a) CPD1: % of teachers who took part in training on student assessment practices in the past 12 months. b) CPD2: % involved in training on analysis and use of student assessment results in the same period.
**Purpose:** Cross-national assessment-training and assessment-practice findings. **Visual forms:** Scatter.
**Physical source pages:** 12.
[PNG](<06_Main_Findings/P122-Fig11-Teachers%20trained%20on%20assessment%20during%20their%20ITE%20and%20their%20CPD%20(%25),%202018.%20Note%20Authors%E2%80%99%20elaboration%20on%20TALIS%20data.%20The%20horizontal%20and%20vertical%20dashed%20lines%20show%20the%20median%20values....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=360>) · Original paper and literature note are not hosted.
### P122 - Fig. 12
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 12. Teachers trained on assessment either during ITE or CPD and declaring a high need for continuous professional development (CPD) on assessment (%), 2018. Note: Authors’ elaboration on TALIS data. The horizontal and vertical dashed lines show the median values of the training need and ITE/CPD indicators, respectively.
**Purpose:** Cross-national assessment-training and assessment-practice findings. **Visual forms:** Scatter.
**Physical source pages:** 13.
[PNG](<06_Main_Findings/P122-Fig12-Teachers%20trained%20on%20assessment%20either%20during%20ITE%20or%20CPD%20and%20declaring%20a%20high%20need%20for%20continuous%20professional%20development%20(CPD)%20on%20assessment%20(%25),%202018.%20Note%20Authors%E2%80%99....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=361>) · Original paper and literature note are not hosted.
### P122 - Fig. 13
**Paper:** Preparing assessment literate teachers: A cross-national trend analysis
**Original caption:** Fig. 13. Teachers trained on assessment during their ITE/CPD and a) giving immediate feedback on tasks b) letting students evaluate their own progress c) administering their own assessment d) providing also written feedback on student work (%), 2018. Note: Authors’ elaboration on TALIS data. The horizontal and vertical dashed lines show the median values of the assessment practices and ITE/CPD indicators, respectively.
**Purpose:** Relationships between teacher training and four assessment practices. **Visual forms:** Scatter, Composite.
**Physical source pages:** 13.
[PNG](<06_Main_Findings/P122-Fig13-Teachers%20trained%20on%20assessment%20during%20their%20ITE%20CPD%20and%20a)%20giving%20immediate%20feedback%20on%20tasks%20b)%20letting%20students%20evaluate%20their%20own%20progress%20c)%20administering%20their%20own....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=362>) · Original paper and literature note are not hosted.
### P123 - Fig. 5
**Paper:** Professional Development of New Higher Education Teachers With Information and Communication Technology in Shanghai: A Kirkpatrick’s Evaluation Approach
**Original caption:** Figure 5. Major themes of opinions received from new HE teachers six months after the ICT module. Note. A1–1. LMS supported blended learning; A1–2. Cutting-edge educational technology; A1–3. Free online educational resources; A1–4. Mind mapping tool as advanced organizer; A1–5. PowerPoint for presentation; A1–6. ICT-enhanced instructional design; A2–1. Social media supported collaborative learning; A2–2. Cognitive tool supported critical thinking process; A2–3. ICT-facilitated learning engagement; A2–4. Peer evaluation of online learning products; A3–1. Prolong the time for some important topics; A3–2. Continuing professional development; A3–3. Experienced tutor guidance and knowledge sharing A3–4. Better university-level policy and ICT infrastructure; A3–5. ICT-supported learning assessment strategies.
**Purpose:** Summarizes themes reported by teachers and department heads after the training. **Visual forms:** Line/profile plot.
**Physical source pages:** 18.
[PNG](<06_Main_Findings/P123-Fig05-Major%20themes%20of%20opinions%20received%20from%20new%20HE%20teachers%20six%20months%20after%20the%20ICT%20module.%20Note.%20A1-1.%20LMS%20supported%20blended%20learning;%20A1-2.%20Cutting-edge%20educational%20technology;....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=363>) · Original paper and literature note are not hosted.
### P123 - Fig. 6
**Paper:** Professional Development of New Higher Education Teachers With Information and Communication Technology in Shanghai: A Kirkpatrick’s Evaluation Approach
**Original caption:** Figure 6. Major themes of opinions received from department heads. Note. P1. Higher ICT capacity; P2. Open and positive attitudes towards ICT application; P3. Active in innovative teaching practice; P4. Students were more engaged; P5. Students were satisfied with the new teachers’ courses; P6. Willing to communicate and share experience; N1. Lack of sustained professional development; N2. Needs more attention on the classroom and students; N3. The role of ICT can be further explored; N4. Lack of ICT diffusion in the whole teacher community.
**Purpose:** Summarizes themes reported by teachers and department heads after the training. **Visual forms:** Line/profile plot.
**Physical source pages:** 21.
[PNG](<06_Main_Findings/P123-Fig06-Major%20themes%20of%20opinions%20received%20from%20department%20heads.%20Note.%20P1.%20Higher%20ICT%20capacity;%20P2.%20Open%20and%20positive%20attitudes%20towards%20ICT%20application;%20P3.%20Active%20in%20innovative....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=364>) · Original paper and literature note are not hosted.
### P124 - Fig. 1
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** Fig. 1. Pooled data analysis.
**Purpose:** Pooled associations of professional learning communities with teacher outcomes. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P124-Fig01-Pooled%20data%20analysis.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=365>) · Original paper and literature note are not hosted.
### P127 - Fig. 4
**Paper:** Teacher evaluation – A resource for teaching quality or a burden? An analysis using TALIS 2018 data
**Original caption:** Fig. 4. Effects of principal-led and teacher-led teacher evaluations on teaching quality and well-being (Model fit: RMSEA = .06; CFI = .92; TLI = .90; n = 103098; all paths p < .001; dashed path is non-significant).
**Purpose:** Primary estimated teacher-evaluation associations. **Visual forms:** Path diagram.
**Physical source pages:** 5.
[PNG](<06_Main_Findings/P127-Fig04-Effects%20of%20principal-led%20and%20teacher-led%20teacher%20evaluations%20on%20teaching%20quality%20and%20well-being%20(Model%20fit%20RMSEA%20=%20.06;%20CFI%20=%20.92;%20TLI%20=%20.90;%20n%20=%20103098;%20all%20paths%20p%20less%20than....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=366>) · Original paper and literature note are not hosted.
### P131 - Fig. 3
**Paper:** Educational equity measurement: A comparative analysis of equity indicators
**Original caption:** Fig. 3. Standardized Root Mean Square Error (RMSE) Averaged Across 37 Countries by Sampling Percentage (25%–40%). Note. ESCS R² and slope indicate the coefficient of determination and slope in the regression model of ESCS; CI = the concentration index; IOpI = the inequality of opportunity index.
**Purpose:** Main simulation comparison of equity-estimation accuracy by sample size. **Visual forms:** Line/profile plot.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P131-Fig03-Standardized%20Root%20Mean%20Square%20Error%20(RMSE)%20Averaged%20Across%2037%20Countries%20by%20Sampling%20Percentage%20(25%25-40%25).%20Note.%20ESCS%20R2%20and%20slope%20indicate%20the%20coefficient%20of%20determination%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=367>) · Original paper and literature note are not hosted.
### P140 - Fig. 1
**Paper:** School productive performance and technology gaps: New evidence from PISA 2018
**Original caption:** Fig. 1. The distribution of efficiency scores under country frontier technology.
**Purpose:** Main cross-country educational efficiency estimates. **Visual forms:** Density/histogram/boxplot.
**Physical source pages:** 14.
[PNG](<06_Main_Findings/P140-Fig01-The%20distribution%20of%20efficiency%20scores%20under%20country%20frontier%20technology.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=368>) · Original paper and literature note are not hosted.
### P140 - Fig. 2
**Paper:** School productive performance and technology gaps: New evidence from PISA 2018
**Original caption:** Fig. 2. Relationship between efficiency scores, maths and reading performance.
**Purpose:** Estimated efficiency related to mathematics and reading performance. **Visual forms:** Scatter, Composite.
**Physical source pages:** 15.
[PNG](<06_Main_Findings/P140-Fig02-Relationship%20between%20efficiency%20scores,%20maths%20and%20reading%20performance.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=369>) · Original paper and literature note are not hosted.
### P140 - Fig. 3
**Paper:** School productive performance and technology gaps: New evidence from PISA 2018
**Original caption:** Fig. 3. Kernel densities for efficiency, meta-efficiency scores and technological gap.
**Purpose:** Comparison of efficiency, meta-efficiency and technology-gap estimates. **Visual forms:** Density/histogram/boxplot.
**Physical source pages:** 16.
[PNG](<06_Main_Findings/P140-Fig03-Kernel%20densities%20for%20efficiency,%20meta-efficiency%20scores%20and%20technological%20gap.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=370>) · Original paper and literature note are not hosted.
### P140 - Fig. 4
**Paper:** School productive performance and technology gaps: New evidence from PISA 2018
**Original caption:** Fig. 4. The distribution of technology gap.
**Purpose:** Main technology-gap comparison across countries. **Visual forms:** Density/histogram/boxplot.
**Physical source pages:** 16.
[PNG](<06_Main_Findings/P140-Fig04-The%20distribution%20of%20technology%20gap.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=371>) · Original paper and literature note are not hosted.
### P142 - Fig. 1
**Paper:** Active Fairness in Algorithmic Decision Making
**Original caption:** Figure 1: Achieving calibration and single error rate parity: classifiers with group-level information budgets vs. naive randomization. Rows correspond to analysis on two different datasets: the Mexican poverty and adult income datasets. Green and yellow colors correspond to error rates for two population subgroups (e.g., white and non-white individuals). Solid black lines represent the space of possible calibrated classifiers, from the perfect classifier (at the origin), to the trivial classifier that assigns the base rate probability to all individuals (intersection with the grey line). Panels in the first column (A and D) show the generalized false positive and false negative rates (GFPR and GFNR [21]) of classifiers that randomize an increasing proportion of individuals (0% to 100%). Dashed lines connect classifiers that achieve single error parity constraints. In line with [21], naive randomization is able to achieve calibration and any single error parity constraint. Panels in the second column (B and E) show results for classifiers with increasing group-level budgets (i.e., increasing feature sets), showing that these are also effective in achieving parity on either false positives, false negatives, or equal cost, while maintaining calibration; yet without resorting to naive randomization. Finally, panels in the third column (C and F) compare the efficiency of both methods in achieving equal opportunity and calibration, under an information budget restriction ( ̄b < 20% in C, and ̄b < 5% in F). Classifiers with group-level budgets Pareto-dominate randomized classifiers by a wide margin, i.e. for the same information budget, both population subgroups are better off, incurring substantially lower false positive and false negative errors.
**Purpose:** Primary empirical comparison of fairness and information-efficiency achieved by classifiers. **Visual forms:** Composite, Trade-off curve.
**Physical source pages:** 3.
[PNG](<06_Main_Findings/P142-Fig01-Achieving%20calibration%20and%20single%20error%20rate%20parity%20classifiers%20with%20group-level%20information%20budgets%20vs.%20naive%20randomization.%20Rows%20correspond%20to%20analysis%20on%20two%20different....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=372>) · Original paper and literature note are not hosted.
### P142 - Fig. 2
**Paper:** Active Fairness in Algorithmic Decision Making
**Original caption:** Figure 2: Achieving parity in false positives and false negatives (equal odds) via group-level information budgets. Results correspond to the Mexican poverty dataset. Achievable regions of classifiers for each population subgroup are plotted in blue and yellow. The outer and inner FPR-FNR curves of each achievable region correspond to classifiers using maximum and minimum information budgets. Points along the curve correspond to different values of the decision threshold. It is observed that active classifiers with group-level budgets achieve parity in both FNR and FPR (equal odds). Moreover, they provide equal odds solutions anywhere on the overlap of the achievable regions for both subgroups and thus along the entire FNR-FPR trade-off.
**Purpose:** Primary empirical comparison of fairness and information-efficiency achieved by classifiers. **Visual forms:** Trade-off curve.
**Physical source pages:** 5.
[PNG](<06_Main_Findings/P142-Fig02-Achieving%20parity%20in%20false%20positives%20and%20false%20negatives%20(equal%20odds)%20via%20group-level%20information%20budgets.%20Results%20correspond%20to%20the%20Mexican%20poverty%20dataset.%20Achievable%20regions....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=373>) · Original paper and literature note are not hosted.
### P142 - Fig. 3
**Paper:** Active Fairness in Algorithmic Decision Making
**Original caption:** Figure 3: Achievable space of equal odds classifiers. Comparison of active classifiers with group-level (green) and individuallevel (yellow) information budgets, and naive randomization (blue) [11]. Classifiers are constrained by an information budget ̄b ≤ bmax . Figure A corresponds to Mexican poverty, B to German credit and C to Heart health under different budget constraints. It is observed that active classifiers yield equal odds solutions along the FPR-FNR trade-off, without resorting to randomization. Moreover, in budget-constraint settings, the active classifiers are substantially more information-efficient. The individuallevel budget Pareto-dominates the group-level classifiers which in turn dominate randomized classifiers.
**Purpose:** Primary empirical comparison of fairness and information-efficiency achieved by classifiers. **Visual forms:** Composite, Trade-off curve.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P142-Fig03-Achievable%20space%20of%20equal%20odds%20classifiers.%20Comparison%20of%20active%20classifiers%20with%20group-level%20(green)%20and%20individuallevel%20(yellow)%20information%20budgets,%20and%20naive%20randomization....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=374>) · Original paper and literature note are not hosted.
### P143 - Fig. 1
**Paper:** Adapting Fairness Interventions to Missing Values
**Original caption:** Figure 1: Comparison of adaptive fairness-intervention algorithms on IPUMS Adult and COMPAS using FairProjection. For each original dataset, we generate three datasets with missing values corresponding to MCAR, MAR, and MNAR mechanisms (Appendix D.2). Error bars depict the standard error of 10 runs with different train-test splits.
**Purpose:** Primary fairness–accuracy comparisons under three missingness mechanisms. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P143-Fig01-Comparison%20of%20adaptive%20fairness-intervention%20algorithms%20on%20IPUMS%20Adult%20and%20COMPAS%20using%20FairProjection.%20For%20each%20original%20dataset,%20we%20generate%20three%20datasets%20with%20missing%20values....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=375>) · Original paper and literature note are not hosted.
### P143 - Fig. 2
**Paper:** Adapting Fairness Interventions to Missing Values
**Original caption:** Figure 2: Comparison of adaptive fairness-intervention algorithms on HSLS dataset, using DispMistreatment (left) and FairProjection (right). Error bars depict the standard error of 10 runs with different train-test splits.
**Purpose:** Main HSLS fairness–accuracy comparisons. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 9.
[PNG](<06_Main_Findings/P143-Fig02-Comparison%20of%20adaptive%20fairness-intervention%20algorithms%20on%20HSLS%20dataset,%20using%20DispMistreatment%20(left)%20and%20FairProjection%20(right).%20Error%20bars%20depict%20the%20standard%20error%20of%2010....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=376>) · Original paper and literature note are not hosted.
### P143 - Fig. 3
**Paper:** Adapting Fairness Interventions to Missing Values
**Original caption:** Figure 3: Results for FairMissBag on HSLS, using Reduction, EqOdds, and FairProjection. Fairness-accuracy curves under mean, KNN, and iterative imputation are shown, respectively. Error bars depict the standard error of 5 runs with different train-test splits.
**Purpose:** Main ensemble-method performance comparison under imputation methods. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 9.
[PNG](<06_Main_Findings/P143-Fig03-Results%20for%20FairMissBag%20on%20HSLS,%20using%20Reduction,%20EqOdds,%20and%20FairProjection.%20Fairness-accuracy%20curves%20under%20mean,%20KNN,%20and%20iterative%20imputation%20are%20shown,%20respectively.%20Error....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=377>) · Original paper and literature note are not hosted.
### P144 - Fig. 1
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure 1: AI Adoption and Professional Development over Time Notes: This figure reports teacher AI adoption and teacher AI professional development trends for 2022–2025. Panel (a) reports the share of principals who replied “yes” to whether “Teachers at my school were using Generative AI tools for teaching-related tasks on or before” June 2022, June 2023, June 2024, and June 2025. Panel (b) reports the share responding “yes” to whether “Staff at my school had received any professional development training on generative AI” by those same June dates. Both panels stratify the series by economic disadvantage, which is based on the free/reduced-price lunch share of students split at the median. All means are IPW-weighted adoption shares by year. The stars reported above each pair of points correspond to the significance level of a difference in means test between high- and low-disadvantage school rates in a given year: ∗p < 0.10, ∗∗ p < 0.05,∗∗∗ p < 0.01.
**Purpose:** AI adoption and training trends are primary empirical findings. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P144-Fig01-AI%20Adoption%20and%20Professional%20Development%20over%20Time%20Notes%20This%20figure%20reports%20teacher%20AI%20adoption%20and%20teacher%20AI%20professional%20development%20trends%20for%202022-2025.%20Panel%20(a)%20reports....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=378>) · Original paper and literature note are not hosted.
### P144 - Fig. 2
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure 2: Student and Educator Use of AI Notes: This figure reports principal responses about generative AI use by students and educators. Panel (a) reports principal responses to the question: “When AI tools are used in class, how frequently do they replace direct wholeclass instruction?” Panel (b) reports principal responses to the question: “What percent of students at your school do you think use Generative AI for the following tasks?” Tasks include homework help and worked examples; essay drafting and writing feedback; brainstorming and project ideation; personalized tutoring; and study guides, quizzes, and flashcards. Panel (c) reports principal responses to the question: “What percent of teachers at your school do you think use Generative AI for the following tasks?” Tasks include any use, lesson planning, grading, detecting academic-integrity violations, assessment and feedback, and administrative tasks. Panel (d) reports the share of principals answering Yes” to the question: “Do you use Generative AI for any of the following tasks?” Tasks include offloading administrative busywork; streamlining data analysis and reporting; designing professional development; and prepping meetings and presentations. Estimates are computed among principals with non-missing responses to the relevant item and are IPW-weighted to be nationally representative.
**Purpose:** Primary patterns of student and educator use. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 11.
[PNG](<06_Main_Findings/P144-Fig02-Student%20and%20Educator%20Use%20of%20AI%20Notes%20This%20figure%20reports%20principal%20responses%20about%20generative%20AI%20use%20by%20students%20and%20educators.%20Panel%20(a)%20reports%20principal%20responses%20to%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=379>) · Original paper and literature note are not hosted.
### P144 - Fig. B.1
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure B.1: Teacher AI Adoption and Professional Development Timelines Notes: This figure documents the rapid diffusion of generative AI in U.S. K–12 schools between 2022 and 2025. Panel (a) plots the IPW-weighted share of schools for which principals reported that teachers were using Generative AI tools for teaching-related tasks on or before June of each year, 2022–2025. Panel (b) plots the IPW-weighted share of schools for which principals reported that staff had received any professional development training on generative AI by the same June dates. Both panels show annual point estimates connected by lines on the full sample of 1,254 responding principals; no additional controls or fixed effects are included. The outcomes are binary indicators for any reported teacher AI adoption and any reported AI professional development, respectively. All estimates are IPW-weighted to be nationally representative of U.S. K–12 schools.
**Purpose:** Teacher adoption and professional-development trends. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 32.
[PNG](<06_Main_Findings/P144-FigB.1-Teacher%20AI%20Adoption%20and%20Professional%20Development%20Timelines%20Notes%20This%20figure%20documents%20the%20rapid%20diffusion%20of%20generative%20AI%20in%20U.S.%20K-12%20schools%20between%202022%20and%202025.%20Panel....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=380>) · Original paper and literature note are not hosted.
### P145 - Fig. 1
**Paper:** AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting
**Original caption:** Fig. 1. Comparison of learning gains: A comparison of mean post-test performance between students taught with the in-class active learning and students taught with the AI tutor. Dotted line represents students’ mean baseline knowledge before the lesson (i.e., the pre-test scores of both groups). Error bars show one standard error of the mean.
**Purpose:** Primary randomized comparison of learning gains. **Visual forms:** Bar chart.
**Physical source pages:** 3.
[PNG](<06_Main_Findings/P145-Fig01-Comparison%20of%20learning%20gains%20A%20comparison%20of%20mean%20post-test%20performance%20between%20students%20taught%20with%20the%20in-class%20active%20learning%20and%20students%20taught%20with%20the%20AI%20tutor.%20Dotted....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=381>) · Original paper and literature note are not hosted.
### P145 - Fig. 3
**Paper:** AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting
**Original caption:** Fig. 3. Student perception of learning experiences: Level of agreement to statements about perceptions of learning experiences, comparing students taught by in-class active learning and students taught with the AI tutor. Error bars show 1 standard error of the mean. Asterisks above the bars denote P-values generated by dependent t-tests (***p<0.001).
**Purpose:** Primary comparison of perceived learning experiences. **Visual forms:** Bar chart.
**Physical source pages:** 5.
[PNG](<06_Main_Findings/P145-Fig03-Student%20perception%20of%20learning%20experiences%20Level%20of%20agreement%20to%20statements%20about%20perceptions%20of%20learning%20experiences,%20comparing%20students%20taught%20by%20in-class%20active%20learning%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=382>) · Original paper and literature note are not hosted.
### P149 - Fig. 3
**Paper:** A systematic review and meta-analysis of the evidence on learning during the COVID-19 pandemic
**Original caption:** Fig. 3 | Forest plot showing individual estimates by study (n = 42), averaged across subjects and grade levels, and the overall effect size estimate, pooled using inverse variance weights and a random-effects model. Effect sizes are expressed in standard deviations, using Cohen’s d, with 95% CI, and are sorted by magnitude.
**Purpose:** Pooled primary learning-deficit estimate. **Visual forms:** Forest plot.
**Physical source pages:** 4.
[PNG](<06_Main_Findings/P149-Fig03-Forest%20plot%20showing%20individual%20estimates%20by%20study%20(n%20=%2042),%20averaged%20across%20subjects%20and%20grade%20levels,%20and%20the%20overall%20effect%20size%20estimate,%20pooled%20using%20inverse%20variance....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=383>) · Original paper and literature note are not hosted.
### P152 - Fig. 2
**Paper:** Domain Adaptation under Missingness Shift
**Original caption:** Figure 2: MSE/Var(Y) of linear models on (a) synthetic and (b) semisynthetic data across varying mˢ and mᵗ. (a) Target domain error of linear models vs. ε. Oracle & closed-form overlap. (b) Target domain error of linear models as the L2-norm between mˢ and mᵗ varies. Best-fit line with 95% confidence intervals from bootstrapping.
**Purpose:** Primary synthetic and semi-synthetic performance comparison. **Visual forms:** Composite, Line plot.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P152-Fig02-MSE%20Var(Y)%20of%20linear%20models%20on%20(a)%20synthetic%20and%20(b)%20semisynthetic%20data%20across%20varying%20ms%20and%20mt.%20(a)%20Target%20domain%20error%20of%20linear%20models%20vs.%20%CE%B5.%20Oracle%20&%20closed-form%20overlap....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=384>) · Original paper and literature note are not hosted.
### P154 - Fig. 2
**Paper:** EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE
**Original caption:** Figure 2: Image inpainting example with MNIST dataset using Partial VAE with four settings.
**Purpose:** Qualitative MNIST inpainting comparison across model settings. **Visual forms:** Composite.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P154-Fig02-Image%20inpainting%20example%20with%20MNIST%20dataset%20using%20Partial%20VAE%20with%20four%20settings.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=385>) · Original paper and literature note are not hosted.
### P154 - Fig. 3
**Paper:** EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE
**Original caption:** Figure 3: Information curves of active variable selection, demonstrated on three UCI datasets (based on PNP parameterization of Partial VAE). This displays negative test RMSE (y axis, the lower the better) during the course of active selection (x-axis). Error bars represent standard errors over 10 runs.
**Purpose:** Active-variable acquisition performance on three UCI datasets. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P154-Fig03-Information%20curves%20of%20active%20variable%20selection,%20demonstrated%20on%20three%20UCI%20datasets%20(based%20on%20PNP%20parameterization%20of%20Partial%20VAE).%20This%20displays%20negative%20test%20RMSE%20(y%20axis,%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=386>) · Original paper and literature note are not hosted.
### P154 - Fig. 4
**Paper:** EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE
**Original caption:** Figure 4: First four decision steps on Boston Housing test data. EDDI is “personalized” comparing SING. Full names of the variables are listed in the Appendix B.2.
**Purpose:** Illustration of personalized variable selection on Boston Housing. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P154-Fig04-First%20four%20decision%20steps%20on%20Boston%20Housing%20test%20data.%20EDDI%20is%20%E2%80%9Cpersonalized%E2%80%9D%20comparing%20SING.%20Full%20names%20of%20the%20variables%20are%20listed%20in%20the%20Appendix%20B.2.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=387>) · Original paper and literature note are not hosted.
### P154 - Fig. 5
**Paper:** EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE
**Original caption:** Figure 5: Comparison of DRAL (Lewenberg et al., 2017) and EDDI on Boston Housing dataset. EDDI outperforms DRAL significantly regarding test RMSE in every step.
**Purpose:** Performance comparison with DRAL. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P154-Fig05-Comparison%20of%20DRAL%20(Lewenberg%20et%20al.,%202017)%20and%20EDDI%20on%20Boston%20Housing%20dataset.%20EDDI%20outperforms%20DRAL%20significantly%20regarding%20test%20RMSE%20in%20every%20step.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=388>) · Original paper and literature note are not hosted.
### P154 - Fig. 6
**Paper:** EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE
**Original caption:** Figure 6: Information curves (based on Bernoulli negative log likelihood) of active variable selection on risk assessment task on MIMIC III with PNP setting.
**Purpose:** Active acquisition in MIMIC III risk assessment. **Visual forms:** Line/profile plot.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P154-Fig06-Information%20curves%20(based%20on%20Bernoulli%20negative%20log%20likelihood)%20of%20active%20variable%20selection%20on%20risk%20assessment%20task%20on%20MIMIC%20III%20with%20PNP%20setting.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=389>) · Original paper and literature note are not hosted.
### P154 - Fig. 7
**Paper:** EDDI: Efficient Dynamic Discovery of High-Value Information with Partial VAE
**Original caption:** Figure 7: Information curves of active (grouped) variable selection on risk assessment task on NHANES with PNP setting.
**Purpose:** Grouped active acquisition in NHANES risk assessment. **Visual forms:** Line/profile plot.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P154-Fig07-Information%20curves%20of%20active%20(grouped)%20variable%20selection%20on%20risk%20assessment%20task%20on%20NHANES%20with%20PNP%20setting.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=390>) · Original paper and literature note are not hosted.
### P155 - Fig. 6
**Paper:** Educational participation, cultural resources, and social inequality: Have social inequalities in reading literacy increased or decreased between 2000 and 2022?
**Original caption:** Figure 6. Correlation between socio-economic status (std. ISEI) and reading skills (OLS-regression coefficients) and test for significant changes over time (Chi²-Test). Source. PISA 2000–2022 (imputed and clustered data; own calculation). Notes. This figure shows standardized ISEI coefficients from OLS-regressions with reading skills as the dependent variable. The chi²-values indicate whether the coefficients differ significantly between the respective time points. PISA: Programme for International Student Assessment; ISEI: International Socio-Economic Index of Occupational Status.
**Purpose:** Main reading-inequality gradient over eight survey years. **Visual forms:** Line/profile plot.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P155-Fig06-Correlation%20between%20socio-economic%20status%20(std.%20ISEI)%20and%20reading%20skills%20(OLS-regression%20coefficients)%20and%20test%20for%20significant%20changes%20over%20time%20(Chi2-Test).%20Source.%20PISA....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=391>) · Original paper and literature note are not hosted.
### P156 - Fig. 3
**Paper:** Effects of LLM use and note-taking on reading comprehension and memory: A randomised experiment in secondary schools
**Original caption:** Fig. 3. Distribution of test performance by condition and group for comprehension, literal retention, and free recall. Mean values are indicated by the two large circles within each facet, whereas the smaller points show individual students’ scores. Error bars indicate one standard error above and below the mean. Group 1 is shown on the left facet of each plot, comparing LLM (red) and Notes (blue). Group 2 is on the right facet of each plot, comparing LLM (red) and LLM + Notes (green). Comprehension (left, max 12 points): Group 1 - LLM: M = 4.00 (SD = 2.44), Notes: M = 4.89 (SD = 2.52); Group 2 - LLM: M = 3.80 (SD = 2.47), LLM + Notes: M = 4.11 (SD = 2.65). Literal retention (middle, max 20 points): Group 1 - LLM: M = 8.83 (SD = 3.96), Notes: M = 10.80 (SD = 4.29); Group 2 - LLM: M = 8.95 (SD = 4.29), LLM + Notes: M = 9.68 (SD = 4.83). Free recall (right, max 50 points): Group 1 - LLM: M = 4.32 (SD = 4.15), Notes: M = 5.36 (SD = 5.49); Group 2 - LLM: M = 4.32 (SD = 4.63), LLM + Notes: M = 4.20 (SD = 5.07). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
**Purpose:** Primary comprehension and memory outcome comparison. **Visual forms:** Composite, Distribution plot.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P156-Fig03-Distribution%20of%20test%20performance%20by%20condition%20and%20group%20for%20comprehension,%20literal%20retention,%20and%20free%20recall.%20Mean%20values%20are%20indicated%20by%20the%20two%20large%20circles%20within%20each....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=392>) · Original paper and literature note are not hosted.
### P156 - Fig. 4
**Paper:** Effects of LLM use and note-taking on reading comprehension and memory: A randomised experiment in secondary schools
**Original caption:** Fig. 4. Differences in student engagement by group and condition. The top and middle panels show ratings for emotional and cognitive engagement measures, respectively, on a 1–5 scale. The bottom panel displays time on task, a behavioural engagement measure, in minutes. Each point represents the mean for a condition, with error bars indicating one standard error above and below the mean.
**Purpose:** Primary student engagement differences between conditions. **Visual forms:** Composite, Point and interval plot.
**Physical source pages:** 13.
[PNG](<06_Main_Findings/P156-Fig04-Differences%20in%20student%20engagement%20by%20group%20and%20condition.%20The%20top%20and%20middle%20panels%20show%20ratings%20for%20emotional%20and%20cognitive%20engagement%20measures,%20respectively,%20on%20a%201-5%20scale....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=393>) · Original paper and literature note are not hosted.
### P158 - Fig. 6
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 6 Forest plot of Hedges’ g effect sizes with 95% confidence intervals from the the random-effects meta-analysis and the μ with credible interval (CrI) from the robust bayesian meta-analysis
**Purpose:** Primary individual and pooled cognitive learning effects. **Visual forms:** Forest plot.
**Physical source pages:** 25.
[PNG](<06_Main_Findings/P158-Fig06-Forest%20plot%20of%20Hedges%E2%80%99%20g%20effect%20sizes%20with%2095%25%20confidence%20intervals%20from%20the%20the%20random-effects%20meta-analysis%20and%20the%20%CE%BC%20with%20credible%20interval%20(CrI)%20from%20the%20robust%20bayesian....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=394>) · Original paper and literature note are not hosted.
### P158 - Fig. 12
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 12 Summary of practical takeaways for educators based on the findings in this work
**Purpose:** Synthesizes evidence-based practical takeaways. **Visual forms:** Conceptual evidence synthesis.
**Physical source pages:** 43.
[PNG](<06_Main_Findings/P158-Fig12-Summary%20of%20practical%20takeaways%20for%20educators%20based%20on%20the%20findings%20in%20this%20work.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=395>) · Original paper and literature note are not hosted.
### P159 - Fig. 2
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure 2: The Impact of AI Access on Generative AI Usage During the Learning Phase. Notes: This figure shows the impact of AI access on generative AI usage during the learning phase in Session One. Gray bars represent control-group means, while red outlines show treatment effects. The first measure is constructed from activity in the ChatGPT account provided to each student. The remaining three measures are based on self-reported usage collected in the exit survey. Vertical bars represent 95 percent confidence intervals.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 29.
[PNG](<06_Main_Findings/P159-Fig02-The%20Impact%20of%20AI%20Access%20on%20Generative%20AI%20Usage%20During%20the%20Learning%20Phase.%20Notes%20This%20figure%20shows%20the%20impact%20of%20AI%20access%20on%20generative%20AI%20usage%20during%20the%20learning%20phase%20in....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=396>) · Original paper and literature note are not hosted.
### P159 - Fig. 3
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure 3: Types of Generative AI Use During the Learning Phase. Panel A. Self-reported usage. Panel B. ChatGPT logs. Notes: This figure shows AI usage types among treated students. Panel A presents self-reported usage types collected in the exit survey; students could select multiple categories. Panel B is constructed from the actual ChatGPT conversation logs. We first use an LLM to classify each student prompt into one of six categories: explaining concepts, summarizing readings, drafting responses, proofreading, editing responses, and other uses (see Appendix B.6 for the classification procedure). We then construct studentlevel indicators: for each category, the indicator equals one if any of the student’s prompts fell into that category. Students who were assigned to the AI-allowed condition but did not use the provided ChatGPT account are coded as zero for all categories.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 30.
[PNG](<06_Main_Findings/P159-Fig03-Types%20of%20Generative%20AI%20Use%20During%20the%20Learning%20Phase.%20Panel%20A.%20Self-reported%20usage.%20Panel%20B.%20ChatGPT%20logs.%20Notes%20This%20figure%20shows%20AI%20usage%20types%20among%20treated%20students.%20Panel%20A....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=397>) · Original paper and literature note are not hosted.
### P159 - Fig. 4
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure 4: Distributional Effects of AI Access on Test Performance. Panel A. Treatment effects by score threshold. Panel B. Session One (knowledge acquisition). Panel C. Session Two (knowledge retention). Notes: Panel A plots ITT estimates of the effect of AI access on indicators for scoring at or above each threshold, separately for Session One (navy circles) and Session Two (maroon diamonds). Each estimate comes from a regression that includes strata fixed effects and controls selected through a double-lasso procedure (Belloni et al., 2014). Vertical bars show 95 percent confidence intervals based on heteroskedasticityrobust standard errors. Panels B and C show the distribution of test scores by treatment assignment. The x-axis shows the number of correct responses; the y-axis shows the percentage of students achieving each score. The Session One test had five questions; the Session Two test had ten questions.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 31.
[PNG](<06_Main_Findings/P159-Fig04-Distributional%20Effects%20of%20AI%20Access%20on%20Test%20Performance.%20Panel%20A.%20Treatment%20effects%20by%20score%20threshold.%20Panel%20B.%20Session%20One%20(knowledge%20acquisition).%20Panel%20C.%20Session%20Two....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=398>) · Original paper and literature note are not hosted.
### P159 - Fig. 5
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure 5: Effect Sizes Across AI-and-Learning Experiments Notes: This figure compares treatment effect sizes (in standard deviations) across randomized experiments that provide students with AI tools during a learning task and measure performance on unassisted assessments. Marker shapes indicate the outcome domain: diamonds for this paper, circles for math, squares for coding, crosses for language learning, triangles for economics, hollow circles for medicine, and hollow squares for science and engineering. Our estimates include both knowledge test scores and essay quality for Session One (immediate) and Session Two (retention), shown above the dashed separator line. Literature estimates come from the 14 studies described in Appendix B.7 and summarized in Appendix Table B3. Horizontal lines show 95 percent confidence intervals. The shaded band and vertical line show the random-effects grand mean across the other studies and its 95 percent confidence interval, estimated via the DerSimonian and Laird (1986) method after aggregating each study’s estimates by inverse-variance weighting (our own estimates are excluded from the pool). The dashed vertical line marks zero.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 32.
[PNG](<06_Main_Findings/P159-Fig05-Effect%20Sizes%20Across%20AI-and-Learning%20Experiments%20Notes%20This%20figure%20compares%20treatment%20effect%20sizes%20(in%20standard%20deviations)%20across%20randomized%20experiments%20that%20provide%20students....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=399>) · Original paper and literature note are not hosted.
### P159 - Fig. 6
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure 6: Effects of AI Access on Essay Outcomes Notes: This figure presents treatment effects of AI access on AI detection and textual similarity (top panel), overall essay quality (middle panel), and individual quality dimensions (bottom panel), measured in standard deviations. The top-panel measures are the AI-detected fraction (the share of text classified as AI-generated by Pangram 4, standardized in the control group), within-group similarity (the average pairwise cosine similarity within treatment×topic×prompt cells), and reading material similarity (between the student essay and the provided reading material). The overall quality measures are overall quality (the grader’s holistic assessment) and the quality index (averaging the five dimensions below), each averaged across the human and AI graders. The quality dimensions are the five sub-components of a standardized rubric (accuracy of content, use of evidence and examples, relevance to the prompt, organization and structure, and writing style and clarity), each averaged across the human and AI graders. Circles represent Session One effects (essays written with or without AI access); diamonds represent Session Two effects (essays written one week later without AI access). Horizontal lines represent 95 percent confidence intervals. Point estimates underlying each coefficient in raw units are reported in Table 5 and Appendix Table A5. The five quality dimensions split by grader appear in Appendix Figure A7.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 33.
[PNG](<06_Main_Findings/P159-Fig06-Effects%20of%20AI%20Access%20on%20Essay%20Outcomes%20Notes%20This%20figure%20presents%20treatment%20effects%20of%20AI%20access%20on%20AI%20detection%20and%20textual%20similarity%20(top%20panel),%20overall%20essay%20quality....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=400>) · Original paper and literature note are not hosted.
### P159 - Fig. 7
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure 7: Actual and Perceived Treatment Effects on Test Performance Notes: This figure shows actual and perceived treatment effects of AI access on test performance. The first bar shows observed performance: gray is the control mean, and green is the estimated treatment effect. The remaining bars show students’ beliefs about own and others’ performance, separately by treatment group: gray is the believed score without AI and green the perceived AI gain. AI-forbidden and AI-allowed belief bars are group means. Vertical bars represent 95 percent confidence intervals.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 34.
[PNG](<06_Main_Findings/P159-Fig07-Actual%20and%20Perceived%20Treatment%20Effects%20on%20Test%20Performance%20Notes%20This%20figure%20shows%20actual%20and%20perceived%20treatment%20effects%20of%20AI%20access%20on%20test%20performance.%20The%20first%20bar%20shows....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=401>) · Original paper and literature note are not hosted.
### P159 - Fig. 8
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure 8: The Average Narrative About AI’s Effect on Learning Notes: This figure shows the aggregate signed causal graph across the 201 open-ended responses codable as a narrative, constructed following Andre et al. (2026). We code each student’s answer to the question “In your opinion, how does generative AI (e.g., ChatGPT) affect student learning in college? Please explain your reasoning” as a causal graph from AI use (left) to learning (right) through the mechanisms the student names; the figure overlays these graphs. Node size is proportional to the share of students who name a mechanism, and link width to the share drawing the link; we omit mechanisms below 10 percent and links below 2 percent. Green links denote mechanisms described as promoting learning, red links mechanisms that harm it, and navy the mode-of-use node. Appendix C.2 describes the coding and its reliability.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 35.
[PNG](<06_Main_Findings/P159-Fig08-The%20Average%20Narrative%20About%20AI%E2%80%99s%20Effect%20on%20Learning%20Notes%20This%20figure%20shows%20the%20aggregate%20signed%20causal%20graph%20across%20the%20201%20open-ended%20responses%20codable%20as%20a%20narrative,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=402>) · Original paper and literature note are not hosted.
### P159 - Fig. A5
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure A5: Test-Score Effects by Item and Excluding One Item at a Time Notes: This figure decomposes the test-score effects of Table 4 by item. The Session One post-learning test has five items and the Session Two test has ten. Panels A and C report the ITT on the probability of answering each item correctly, in percentage points. Panels B and D report the ITT on the fraction of items answered correctly when one item at a time is excluded. The hollow marker keeps all items and reproduces the fraction-correct estimates in Table 4 (in percentage points), and the dashed line marks that benchmark. Each estimate selects double-lasso controls for its own outcome (Belloni et al., 2014) and includes strata fixed effects. Vertical bars are 95 percent confidence intervals from heteroskedasticityrobust standard errors.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 47.
[PNG](<06_Main_Findings/P159-FigA5-Test-Score%20Effects%20by%20Item%20and%20Excluding%20One%20Item%20at%20a%20Time%20Notes%20This%20figure%20decomposes%20the%20test-score%20effects%20of%20Table%204%20by%20item.%20The%20Session%20One%20post-learning%20test%20has%20five....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=403>) · Original paper and literature note are not hosted.
### P159 - Fig. A6
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure A6: Effects of AI Access on Essay Characteristics, by Dimension Notes: This figure presents treatment effects of AI access on individual essay characteristics, measured in standard deviations. Each variable is standardized to have mean zero and standard deviation one in the control group. Variables are grouped into four categories, indicated by colored vertical bars: length (number of tokens, words, and sentences), ease of reading (sentence length, syllables per word, Flesch– Kincaid grade level, and Flesch Reading Ease score, with difficulty measures reversed so that higher values indicate easier readability), lexical diversity (type–token ratio and hapax proportion), and homogeneity and similarity (within-group cosine similarity and reading material cosine similarity). Circles represent Session One effects; diamonds represent Session Two effects. Horizontal lines represent 95 percent confidence intervals based on heteroskedasticity-robust standard errors.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 48.
[PNG](<06_Main_Findings/P159-FigA6-Effects%20of%20AI%20Access%20on%20Essay%20Characteristics,%20by%20Dimension%20Notes%20This%20figure%20presents%20treatment%20effects%20of%20AI%20access%20on%20individual%20essay%20characteristics,%20measured%20in%20standard....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=404>) · Original paper and literature note are not hosted.
### P159 - Fig. A7
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure A7: Effects of AI Access on Essay Quality, by Dimension and Grader Notes: This figure presents treatment effects of AI access on individual essay quality dimensions, measured in standard deviations. The top panel shows human-graded dimensions: essays were evaluated by independent graders on five rubric dimensions plus an overall quality rating (accuracy of content, use of evidence and examples, relevance to the prompt, organization and structure, writing style and clarity, and overall quality). Human-graded regressions are at the student level, averaging each essay’s scores across its graders. The bottom panel shows the same dimensions as scored by an AI grader (see Appendix B.4). Each dimension is scored on a scale from 0 to 10 and standardized to have mean zero and standard deviation one in the control group. Circles represent Session One effects (essays written with or without AI access); diamonds represent Session Two effects (essays written one week later without AI access). Horizontal lines represent 95 percent confidence intervals based on heteroskedasticity-robust standard errors.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 49.
[PNG](<06_Main_Findings/P159-FigA7-Effects%20of%20AI%20Access%20on%20Essay%20Quality,%20by%20Dimension%20and%20Grader%20Notes%20This%20figure%20presents%20treatment%20effects%20of%20AI%20access%20on%20individual%20essay%20quality%20dimensions,%20measured%20in....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=405>) · Original paper and literature note are not hosted.
### P159 - Fig. A8
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure A8: Distribution of Learning-Phase Duration by Treatment Arm Notes: This figure plots the distribution of the Qualtrics-recorded duration of the Session One learning phase, by whole minute, separately for the AI-forbidden and AI-allowed conditions.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 50.
[PNG](<06_Main_Findings/P159-FigA8-Distribution%20of%20Learning-Phase%20Duration%20by%20Treatment%20Arm%20Notes%20This%20figure%20plots%20the%20distribution%20of%20the%20Qualtrics-recorded%20duration%20of%20the%20Session%20One%20learning%20phase,%20by%20whole....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=406>) · Original paper and literature note are not hosted.
### P159 - Fig. A9
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure A9: Mechanisms: Time Allocation Across Learning Activities Notes: This figure presents treatment effects of AI access on self-reported time allocation across different learning activities, measured as each activity’s share of the student’s total self-reported learning time (in percentage points). Students reported how many minutes they spent on each activity during the 35-minute learning phase; each activity’s share is its minutes divided by the sum of minutes across all activities. The survey labels were: “Writing a first draft of your write-up” (Drafting), “Revising and editing your draft” (Editing), “Taking notes on key concepts and facts” (Note-taking), “Planning your writing task structure and arguments” (Organizing), “Finding information about the topic, searching for relevant sources” (Searching), “Reading and comprehending information about the topic” (Reading), and “Casually browsing the web, unrelated to the topic” and “Other activities” combined (Other). Navy bars represent writing activities, maroon bars represent research activities, and orange bars represent other activities. Bars represent the estimated treatment effect for each activity, with horizontal lines showing 95 percent confidence intervals. Numbers in parentheses indicate the control-group mean share.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 51.
[PNG](<06_Main_Findings/P159-FigA9-Mechanisms%20Time%20Allocation%20Across%20Learning%20Activities%20Notes%20This%20figure%20presents%20treatment%20effects%20of%20AI%20access%20on%20self-reported%20time%20allocation%20across%20different%20learning....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=407>) · Original paper and literature note are not hosted.
### P159 - Fig. A11
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure A11: Distribution of Perceived Treatment Effects on Own Test Score Notes: This figure plots the distribution of students’ perceived treatment effect of AI access on their own Session Two test score, separately for AI-forbidden (control) and AI-allowed (treated) students. The perceived effect is each student’s believed test score with AI access minus their believed score without it, expressed in number of questions answered correctly (for treated students, the believed own score minus the believed counterfactual; for control students, the believed counterfactual minus the believed own score). Values below −5 are bottom-coded at −5. Bin width is 1 and bins are centered on integers. In-figure annotations report the share of each group reporting a negative, zero, or positive perceived effect.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 53.
[PNG](<06_Main_Findings/P159-FigA11-Distribution%20of%20Perceived%20Treatment%20Effects%20on%20Own%20Test%20Score%20Notes%20This%20figure%20plots%20the%20distribution%20of%20students%E2%80%99%20perceived%20treatment%20effect%20of%20AI%20access%20on%20their%20own%20Session....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=408>) · Original paper and literature note are not hosted.
### P159 - Fig. C1
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure C1: Distribution of AI Sentiment Scores. Notes: This figure plots the distribution of compound sentiment scores across students who responded to the open-ended question about how generative AI affects student learning. Scores are computed using the VADER algorithm of Hutto and Gilbert (2014) and range from −1 (most negative) to +1 (most positive). The dashed vertical line marks the mean.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 75.
[PNG](<06_Main_Findings/P159-FigC1-Distribution%20of%20AI%20Sentiment%20Scores.%20Notes%20This%20figure%20plots%20the%20distribution%20of%20compound%20sentiment%20scores%20across%20students%20who%20responded%20to%20the%20open-ended%20question%20about%20how....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=409>) · Original paper and literature note are not hosted.
### P159 - Fig. C2
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure C2: Relationship Between AI Sentiment and AI Usage. Notes: This figure presents the relationship between AI sentiment and AI usage. Panel A restricts to students in the AI-allowed condition and shows the proportion who used AI during Session One (from usage logs). Panel B shows an indicator for using AI for academic purposes frequently or very frequently during the semester, for all respondents. Each point represents the mean outcome for respondents within sentiment score bins. Sentiment scores are compound scores computed using the VADER algorithm of Hutto and Gilbert (2014) applied to responses to the open-ended question about AI and learning. Positive values indicate positive sentiment and negative values indicate negative sentiment. The dashed lines show OLS best-fit lines estimated on the microdata. C.2 Narratives as Causal Graphs We characterize the content of the open-ended responses by coding each one as a causal graph, following Andre et al. (2026), who analyze how households and experts explain macroeconomic events. In their setting respondents explain why inflation rose, and the explanation is coded as a chain of cause-and-effect links running from a driving factor to inflation. We adapt their method in one substantive way: their links are unsigned, because
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Line/profile plot.
**Physical source pages:** 76.
[PNG](<06_Main_Findings/P159-FigC2-Relationship%20Between%20AI%20Sentiment%20and%20AI%20Usage.%20Notes%20This%20figure%20presents%20the%20relationship%20between%20AI%20sentiment%20and%20AI%20usage.%20Panel%20A%20restricts%20to%20students%20in%20the%20AI-allowed....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=410>) · Original paper and literature note are not hosted.
### P159 - Fig. C3
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure C3: Three Example Student Narratives, Coded as Signed Causal Graphs. Panel A. A narrative in which AI helps learning. Panel B. A narrative in which AI harms learning. Panel C. A conditional narrative. Notes: Each panel pairs a student’s verbatim open-ended response (left) with the response coded as a signed causal graph from AI use to learning (right), following Andre et al. (2026), so the mapping from text to graph can be assessed directly. Green links denote mechanisms the student describes as promoting learning, red links mechanisms that harm it, and navy the mode-of-use node.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Composite.
**Physical source pages:** 78.
[PNG](<06_Main_Findings/P159-FigC3-Three%20Example%20Student%20Narratives,%20Coded%20as%20Signed%20Causal%20Graphs.%20Panel%20A.%20A%20narrative%20in%20which%20AI%20helps%20learning.%20Panel%20B.%20A%20narrative%20in%20which%20AI%20harms%20learning.%20Panel%20C.%20A....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=411>) · Original paper and literature note are not hosted.
### P159 - Fig. C4
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure C4: Narrative Clusters. Notes: This figure shows the average causal graph within each narrative cluster; panel headings report each cluster’s share of coded narratives. We form the clusters with agglomerative hierarchical clustering, using average linkage on the Jaccard distance between signed-link sets and choosing the number of clusters to maximize the average silhouette. We show the four clusters with at least eight responses and, within each panel, omit mechanisms below 20 percent and links below 6 percent.
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Composite.
**Physical source pages:** 83.
[PNG](<06_Main_Findings/P159-FigC4-Narrative%20Clusters.%20Notes%20This%20figure%20shows%20the%20average%20causal%20graph%20within%20each%20narrative%20cluster;%20panel%20headings%20report%20each%20cluster%E2%80%99s%20share%20of%20coded%20narratives.%20We%20form%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=412>) · Original paper and literature note are not hosted.
### P159 - Fig. D1
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure D1: The Effect of AI Access on Knowledge With and Without Delegation. Notes: The figure plots the change in knowledge relative to no AI, K∗ (τ, m) − K0, against the tutoring gain, for a student who writes the essay (m = 0) and for the same student when delegating (m = 1), with research and writing effort chosen optimally in each case. The curves are drawn for µ = ωK = ωQ = θr = θw = ϕ = 1. The values are illustrative, so the axes show no scale. The two curves are two counterfactual choices for the same student, holding everything else fixed, not average outcomes for two groups of students. The shaded area between the curves is the loss from delegation, the sum of the last two terms of equation (D8). The darker band is the learning displaced by delegation, which does not depend on the tutoring gain. The lighter band is the gain from tutoring lost by delegating, which grows with the tutoring gain, so the delegation curve rises more slowly and the gap between the curves widens. The braces show the two parts at the largest tutoring gain drawn. At the marked point on the no-AI line, the gain from tutoring equals the two losses from delegation. D.4 The Decision to Delegate Work to the AI The student compares the utility obtained with and without delegation, each evaluated at the effort levels of equation (D6). Substituting those effort levels into the objective
**Purpose:** Displays experimental outcomes, robustness analyses, or coded response patterns discussed in the paper. **Visual forms:** Composite.
**Physical source pages:** 88.
[PNG](<06_Main_Findings/P159-FigD1-The%20Effect%20of%20AI%20Access%20on%20Knowledge%20With%20and%20Without%20Delegation.%20Notes%20The%20figure%20plots%20the%20change%20in%20knowledge%20relative%20to%20no%20AI,%20K%E2%88%97%20(%CF%84,%20m)%20-%20K0,%20against%20the%20tutoring%20gain,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=413>) · Original paper and literature note are not hosted.
### P166 - Fig. 3
**Paper:** Imputation and missing indicators for handling missing data in the development and deployment of clinical prediction models: A simulation study
**Original caption:** Figure 3. Brier Score estimates across development/validation scenarios, imputation methods and missingness mechanisms. The vertical dashed lines represent estimates from the complete data scenario (DA + VA). DY = Y included in the imputation model at development, DȲ = Y omitted from the imputation model at development, VY = Y included in the imputation model at validation, VȲ = Y omitted from the imputation model at validation, DA = complete data available at model development, VA = complete data available at validation.
**Purpose:** Primary simulation comparison of predictive Brier scores. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 9.
[PNG](<06_Main_Findings/P166-Fig03-Brier%20Score%20estimates%20across%20development%20validation%20scenarios,%20imputation%20methods%20and%20missingness%20mechanisms.%20The%20vertical%20dashed%20lines%20represent%20estimates%20from%20the%20complete....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=414>) · Original paper and literature note are not hosted.
### P166 - Fig. 4
**Paper:** Imputation and missing indicators for handling missing data in the development and deployment of clinical prediction models: A simulation study
**Original caption:** Figure 4. Calibration slope estimates across imputation strategies, imputation methods and missingness mechanisms. Vertical dashed lines are placed at 1. DY = Y included in the imputation model at development, DȲ = Y omitted from the imputation model at development, VY = Y included in the imputation model at validation, VȲ = Y omitted from the imputation model at validation, DA = complete data available at model development, VA = complete data available at validation.
**Purpose:** Primary calibration-performance comparison across missingness and imputation strategies. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P166-Fig04-Calibration%20slope%20estimates%20across%20imputation%20strategies,%20imputation%20methods%20and%20missingness%20mechanisms.%20Vertical%20dashed%20lines%20are%20placed%20at%201.%20DY%20=%20Y%20included%20in%20the%20imputation....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=415>) · Original paper and literature note are not hosted.
### P166 - Fig. 5
**Paper:** Imputation and missing indicators for handling missing data in the development and deployment of clinical prediction models: A simulation study
**Original caption:** Figure 5. Parameter estimates across all missingness mechanisms. Missingness is fixed at 50%. DY = Y included in the imputation model at development, DȲ = Y omitted from the imputation model at development, VY = Y included in the imputation model at validation, VȲ = Y omitted from the imputation model at validation, DA = complete data available at the development, VA = complete data available at validation.
**Purpose:** Primary simulation estimates of parameter bias across strategies. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 11.
[PNG](<06_Main_Findings/P166-Fig05-Parameter%20estimates%20across%20all%20missingness%20mechanisms.%20Missingness%20is%20fixed%20at%2050%25.%20DY%20=%20Y%20included%20in%20the%20imputation%20model%20at%20development,%20D%C8%B2%20=%20Y%20omitted%20from%20the%20imputation....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=416>) · Original paper and literature note are not hosted.
### P167 - Fig. 4
**Paper:** Imputation Strategies Under Clinical Presence: Impact on Algorithmic Fairness
**Original caption:** Figure 4: AUC performance gaps ∆ and group-specfic AUCs across scenarios on 100 synthetic experiments. If ∆ < 0, the marginalised group has worse AUC than the
**Purpose:** Reports model performance, prioritization, or observed test-use patterns. **Visual forms:** Line/profile plot.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P167-Fig04-AUC%20performance%20gaps%20%E2%88%86%20and%20group-specfic%20AUCs%20across%20scenarios%20on%20100%20synthetic%20experiments.%20If%20%E2%88%86%20less%20than%200,%20the%20marginalised%20group%20has%20worse%20AUC%20than%20the.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=417>) · Original paper and literature note are not hosted.
### P167 - Fig. 5
**Paper:** Imputation Strategies Under Clinical Presence: Impact on Algorithmic Fairness
**Original caption:** Figure 5: Prioritisation performance gaps ∆ across marginalised groups in MIMIC III experiment. If ∆ > 0, the marginalised group has a larger value of the given metric
**Purpose:** Reports model performance, prioritization, or observed test-use patterns. **Visual forms:** Line/profile plot.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P167-Fig05-Prioritisation%20performance%20gaps%20%E2%88%86%20across%20marginalised%20groups%20in%20MIMIC%20III%20experiment.%20If%20%E2%88%86%20greater%20than%200,%20the%20marginalised%20group%20has%20a%20larger%20value%20of%20the%20given%20metric.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=418>) · Original paper and literature note are not hosted.
### P167 - Fig. 7
**Paper:** Imputation Strategies Under Clinical Presence: Impact on Algorithmic Fairness
**Original caption:** Figure 7: Venn diagram of the population distribution in the three
**Purpose:** Reports model performance, prioritization, or observed test-use patterns. **Visual forms:** Line/profile plot.
**Physical source pages:** 19.
[PNG](<06_Main_Findings/P167-Fig07-Venn%20diagram%20of%20the%20population%20distribution%20in%20the%20three.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=419>) · Original paper and literature note are not hosted.
### P167 - Fig. 8
**Paper:** Imputation Strategies Under Clinical Presence: Impact on Algorithmic Fairness
**Original caption:** Figure 8: Evolution of the number of tests
**Purpose:** Reports model performance, prioritization, or observed test-use patterns. **Visual forms:** Line/profile plot.
**Physical source pages:** 19.
[PNG](<06_Main_Findings/P167-Fig08-Evolution%20of%20the%20number%20of%20tests.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=420>) · Original paper and literature note are not hosted.
### P167 - Fig. 9
**Paper:** Imputation Strategies Under Clinical Presence: Impact on Algorithmic Fairness
**Original caption:** Figure 9: Prioritisation performance gaps ∆ across marginalised groups in MIMIC III experiment for 5% additional care. If ∆ > 0, the marginalised group has a larger
**Purpose:** Reports model performance, prioritization, or observed test-use patterns. **Visual forms:** Line/profile plot.
**Physical source pages:** 23.
[PNG](<06_Main_Findings/P167-Fig09-Prioritisation%20performance%20gaps%20%E2%88%86%20across%20marginalised%20groups%20in%20MIMIC%20III%20experiment%20for%205%25%20additional%20care.%20If%20%E2%88%86%20greater%20than%200,%20the%20marginalised%20group%20has%20a%20larger.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=421>) · Original paper and literature note are not hosted.
### P167 - Fig. 11
**Paper:** Imputation Strategies Under Clinical Presence: Impact on Algorithmic Fairness
**Original caption:** Figure 11: Prioritisation performance gaps ∆ across marginalised groups in MIMIC III
**Purpose:** Reports model performance, prioritization, or observed test-use patterns. **Visual forms:** Line/profile plot.
**Physical source pages:** 23.
[PNG](<06_Main_Findings/P167-Fig11-Prioritisation%20performance%20gaps%20%E2%88%86%20across%20marginalised%20groups%20in%20MIMIC%20III.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=422>) · Original paper and literature note are not hosted.
### P169 - Fig. 1
**Paper:** Informative missingness in electronic health record systems: the curse of knowing
**Original caption:** Fig. 1 Impact of differences in missing data mechanisms when applying a prediction model that was derived under informative missingness using different approaches to handle missing data. Plots are based on 500 randomly sampled observations out of a dataset of 20000 observations. Predicted probabilities are jittered for visual clarity. See main text for a description of the scenarios and details about the methods
**Purpose:** Primary simulation shows predictive calibration changes under missingness-mechanism shifts. **Visual forms:** Composite, Calibration plot.
**Physical source pages:** 4.
[PNG](<06_Main_Findings/P169-Fig01-Impact%20of%20differences%20in%20missing%20data%20mechanisms%20when%20applying%20a%20prediction%20model%20that%20was%20derived%20under%20informative%20missingness%20using%20different%20approaches%20to%20handle%20missing....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=423>) · Original paper and literature note are not hosted.
### P172 - Fig. 2
**Paper:** Is free and reduced-price lunch a valid measure of educational disadvantage?
**Original caption:** Figure 2. 8th grade NSLP enrollment and IRS-validated household income and poverty rates (averaged at the school level for all 8th graders in one California Public School District, 2008-09 through 2013-14) (N=14,000) Note: Sample size has been rounded to comply with U.S. Census Bureau disclosure requirements. Source: Linked school district administrative records, for academic years 2008-09 through 2013-14, and Internal Revenue Service 1040 Tax data, from 2000 through 2013. 0 20 40 60 80 100 1 2 3 4 5 6 7 8 9 10 School % enrolled in free or reduced lunch % currently in poverty (IRS) % ever in poverty, K-8 (IRS) % household income < 1.8*poverty (IRS)
**Purpose:** Compares school-level lunch enrollment with IRS-validated income and poverty measures. **Visual forms:** Composite.
**Physical source pages:** 21.
[PNG](<06_Main_Findings/P172-Fig02-8th%20grade%20NSLP%20enrollment%20and%20IRS-validated%20household%20income%20and%20poverty%20rates%20(averaged%20at%20the%20school%20level%20for%20all%208th%20graders%20in%20one%20California%20Public%20School%20District,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=424>) · Original paper and literature note are not hosted.
### P172 - Fig. None
**Paper:** Is free and reduced-price lunch a valid measure of educational disadvantage?
**Original caption:** Figure 1b. Distribution of IRS-reported household income-to-poverty ratio by 8th grade NSLP enrollment in Oregon schools for academic years 2004-05 through 2013-14 (N= 315,000). Note: Sample size, distributions, and density bandwidths have been adjusted to comply with Census Bureau disclosure requirements. Source: Linked Oregon Department of Education school district administrative records, for academic years 2004-05 through 2013-14, and Internal Revenue Service 1040 Tax data, from 2004 through 2013.
**Purpose:** Research figure visually checked in its article context. **Visual forms:** Composite.
**Physical source pages:** 20.
[PNG](<06_Main_Findings/P172-FigNone-Distribution%20of%20IRS-reported%20household%20income-to-poverty%20ratio%20by%208th%20grade%20NSLP%20enrollment%20in%20Oregon%20schools%20for%20academic%20years%202004-05%20through%202013-14%20(N=%20315,000).%20Note....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=425>) · Original paper and literature note are not hosted.
### P175 - Fig. 1
**Paper:** Machine learning and phone data can improve targeting of humanitarian aid
**Original caption:** Fig. 1 | Comparing Novissi targeting to alternatives. a, b, The performance of phone-based targeting (green) compared with alternative approaches that were feasible (red) and unfeasible (blue) in Togo in 2020. Targeting is evaluated for the actual rural Novissi programme (a), which focused on Togo’s 100 poorest cantons (using a 2020 survey representative of mobile subscribers in the 100 cantons, where PMT is a ground truth for poverty since consumption data was not collected in the phone survey); and a hypothetical nationwide anti-poverty programme (using a national field survey conducted in 2018–2019, where consumption is a ground truth for poverty) (b). The darker bar in each pair indicates recall and precision (left axis), which is equivalent to 1 – exclusion error; the lighter bar in each pair indicates area under the curve (right axis). The bar height represents the point estimate from the full simulation; whiskers show s.d. produced from n = 1,000 bootstrap simulations. The figure highlights a subset of the results contained in Table 1.
**Purpose:** Primary comparison of phone-based aid targeting against alternatives. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 2.
[PNG](<06_Main_Findings/P175-Fig01-Comparing%20Novissi%20targeting%20to%20alternatives.%20a,%20b,%20The%20performance%20of%20phone-based%20targeting%20(green)%20compared%20with%20alternative%20approaches%20that%20were%20feasible%20(red)%20and%20unfeasible....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=426>) · Original paper and literature note are not hosted.
### P175 - Fig. 2
**Paper:** Machine learning and phone data can improve targeting of humanitarian aid
**Original caption:** Fig. 2 | Welfare analysis of different targeting mechanisms. Aggregate social welfare is calculated (assuming CRRA utility) under counterfactual targeting approaches. We assume a fixed budget of US$4 million and a population of 154,238, with an equal transfer size for all beneficiaries. Utility curves for feasible targeting mechanisms are shown in solid lines; infeasible targeting mechanisms are shown in dashed lines. The horizontal dotted line indicates total social welfare for a universal basic income programme that provides (very small) transfers to the entire population; vertical dotted lines indicate the targeting threshold and associated transfer size that maximizes social welfare for each targeting mechanism. a, b, Targeting is evaluated for the Novissi anti-poverty programme in Togo’s 100 poorest cantons (a) and a hypothetical nationwide anti-poverty programme (b).
**Purpose:** Primary modeled welfare comparisons across coverage thresholds. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 3.
[PNG](<06_Main_Findings/P175-Fig02-Welfare%20analysis%20of%20different%20targeting%20mechanisms.%20Aggregate%20social%20welfare%20is%20calculated%20(assuming%20CRRA%20utility)%20under%20counterfactual%20targeting%20approaches.%20We%20assume%20a%20fixed....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=427>) · Original paper and literature note are not hosted.
### P177 - Fig. 1
**Paper:** Missing data and prediction: the pattern submodel
**Original caption:** Fig. 1. Simulation results set of parameters: β0 = 1, β1 = 3, β2 = 1, δ1 = 1, δ3 = 1, P(M1 = 1) = 0.5, ν1 = 1, ν2 = 1, ν1,Y = 1, ν2,Y = 1. The missing data mechanisms missing at random (MAR) and missing not at random (MNAR) were generated under a pattern mixture Y (PMY) and selection model Y formulation. Triangles represent the total prediction error (TPE) summed over all missing data patterns. Circles represent the prediction error (PE) for pattern 1, where there is no missing data. Stars represent the PE for Pattern 2, in which x1 is missing.
**Purpose:** Primary simulation comparison of pattern-specific prediction methods. **Visual forms:** Scatter, Composite.
**Physical source pages:** 11.
[PNG](<06_Main_Findings/P177-Fig01-Simulation%20results%20set%20of%20parameters%20%CE%B20%20=%201,%20%CE%B21%20=%203,%20%CE%B22%20=%201,%20%CE%B41%20=%201,%20%CE%B43%20=%201,%20P(M1%20=%201)%20=%200.5,%20%CE%BD1%20=%201,%20%CE%BD2%20=%201,%20%CE%BD1,Y%20=%201,%20%CE%BD2,Y%20=%201.%20The%20missing%20data%20mechanisms%20missing%20at%20random....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=428>) · Original paper and literature note are not hosted.
### P177 - Fig. 2
**Paper:** Missing data and prediction: the pattern submodel
**Original caption:** Fig. 2. The covariates included in the SPS prediction model include partial pressure of oxygen in the arterial blood (pafi), mean blood pressure (meanbp), white blood count (wblc), albumin (alb), APACHE III respiration score (resp), temperature (temp), heart rate per minute (hrt), bilirubin (bili), creatinine (crea), and sodium (sod). There are 23 patterns present in the SUPPORT data, and missing covariates are denoted with “X”. N is the total number of subjects in each missing data pattern. Pattern submodels (PS), multiple imputation with missingness indicators (MIMI), complete case submodels (CCS), and traditional multiple imputation (MI) methods are all compared. The top two figures are the unweighted pattern specific PE, and the bottom two figures are the pattern specific contribution to the PE in which the partial PE is weighted by the observed proportion of individuals in each pattern.
**Purpose:** Clinical example contrasts prediction errors under original and induced missingness. **Visual forms:** Scatter, Complex table, Composite.
**Physical source pages:** 13.
[PNG](<06_Main_Findings/P177-Fig02-The%20covariates%20included%20in%20the%20SPS%20prediction%20model%20include%20partial%20pressure%20of%20oxygen%20in%20the%20arterial%20blood%20(pafi),%20mean%20blood%20pressure%20(meanbp),%20white%20blood%20count%20(wblc),....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=429>) · Original paper and literature note are not hosted.
### P180 - Fig. 1
**Paper:** On the Treatment of Missing Data in Background Questionnaires in Educational Large-Scale Assessments: An Evaluation of Different Procedures
**Original caption:** FIGURE 1. Bias (raw and in %) for the estimated mean and variance of θ in conditions with a moderate number of items (J = 15), a strong correlation between θ and x₂ (ρ_{θx₂} = .70), and 40% missing data. CD = complete data (only PVs); FCS = fully conditional specification; MIM-PE/MIM-LD = missing indicator method with pairwise estimation or listwise deletion (point estimates are identical); MIM-MI = MIM with nested MI.
**Purpose:** Primary method-performance comparison in simulation or the real PISA example. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 16.
[PNG](<06_Main_Findings/P180-Fig01-Bias%20(raw%20and%20in%20%25)%20for%20the%20estimated%20mean%20and%20variance%20of%20%CE%B8%20in%20conditions%20with%20a%20moderate%20number%20of%20items%20(J%20=%2015),%20a%20strong%20correlation%20between%20%CE%B8%20and%20x2%20(%CF%81_%7B%CE%B8x2%7D%20=%20.70),%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=430>) · Original paper and literature note are not hosted.
### P180 - Fig. 2
**Paper:** On the Treatment of Missing Data in Background Questionnaires in Educational Large-Scale Assessments: An Evaluation of Different Procedures
**Original caption:** FIGURE 2. Bias (raw and in %) for the estimated the mean and variance of x₁ and the covariances between θ, x₁, and x₂ in conditions with a moderate number of items (J = 15), a strong correlation between θ and x₂ (ρ_{θx₂} = .70), and 40% missing data. The double tilde symbol (≈) denotes values outside the plotted range. CD = complete data (only PVs); FCS = fully conditional specification; MIM-PE/MIM-LD = missing indicator method with pairwise estimation or listwise deletion (point estimates are identical); MIM-MI = MIM with nested MI.
**Purpose:** Primary method-performance comparison in simulation or the real PISA example. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 17.
[PNG](<06_Main_Findings/P180-Fig02-Bias%20(raw%20and%20in%20%25)%20for%20the%20estimated%20the%20mean%20and%20variance%20of%20x1%20and%20the%20covariances%20between%20%CE%B8,%20x1,%20and%20x2%20in%20conditions%20with%20a%20moderate%20number%20of%20items%20(J%20=%2015),%20a%20strong....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=431>) · Original paper and literature note are not hosted.
### P180 - Fig. 3
**Paper:** On the Treatment of Missing Data in Background Questionnaires in Educational Large-Scale Assessments: An Evaluation of Different Procedures
**Original caption:** FIGURE 3. Bias (in %) for the estimated covariance between θ and x₁ in conditions with a strong correlation between θ and x₂ (ρ_{θx₁} = .70) and data MARₓ (50%). CD = complete data (only PVs); FCS = fully conditional specification; MIM-PE/MIM-LD = missing indicator method with pairwise estimation or listwise deletion (point estimates are identical); MIM-MI = MIM with nested MI.
**Purpose:** Primary method-performance comparison in simulation or the real PISA example. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 18.
[PNG](<06_Main_Findings/P180-Fig03-Bias%20(in%20%25)%20for%20the%20estimated%20covariance%20between%20%CE%B8%20and%20x1%20in%20conditions%20with%20a%20strong%20correlation%20between%20%CE%B8%20and%20x2%20(%CF%81_%7B%CE%B8x1%7D%20=%20.70)%20and%20data%20MARx%20(50%25).%20CD%20=%20complete%20data%20(only....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=432>) · Original paper and literature note are not hosted.
### P180 - Fig. 4
**Paper:** On the Treatment of Missing Data in Background Questionnaires in Educational Large-Scale Assessments: An Evaluation of Different Procedures
**Original caption:** FIGURE 4. Bias (in %) for the estimated regression coefficients in the regression of θ on x₁ and x₂ in conditions with a strong correlation between θ and x₂ (ρ_{θx₂} = .70) and 40% missing data (MCAR). CD = complete data (only PVs); FCS = fully conditional specification; MIM-PE = missing indicator method with pairwise estimation; MIM-LD = MIM with listwise deletion; MIM-MI = MIM with nested MI.
**Purpose:** Primary method-performance comparison in simulation or the real PISA example. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 19.
[PNG](<06_Main_Findings/P180-Fig04-Bias%20(in%20%25)%20for%20the%20estimated%20regression%20coefficients%20in%20the%20regression%20of%20%CE%B8%20on%20x1%20and%20x2%20in%20conditions%20with%20a%20strong%20correlation%20between%20%CE%B8%20and%20x2%20(%CF%81_%7B%CE%B8x2%7D%20=%20.70)%20and%2040%25....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=433>) · Original paper and literature note are not hosted.
### P180 - Fig. 5
**Paper:** On the Treatment of Missing Data in Background Questionnaires in Educational Large-Scale Assessments: An Evaluation of Different Procedures
**Original caption:** FIGURE 5. Bias (in %) for the estimated correlation coefficients of θ with x₁ and the regression coefficients in the regression of θ on any pair of variables in x₁ and x₃ in conditions with high reliability (J = 32). The point range and symbols indicate the range of and median bias. The double tilde symbol (≈) denotes values outside the plotted range. The individual points represent the bias for each individual parameter. FCS = fully conditional specification; MIM-PE = missing indicator method with pairwise estimation; MIM-LD = MIM with listwise deletion; MIM-MI = MIM with nested MI.
**Purpose:** Primary method-performance comparison in simulation or the real PISA example. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 25.
[PNG](<06_Main_Findings/P180-Fig05-Bias%20(in%20%25)%20for%20the%20estimated%20correlation%20coefficients%20of%20%CE%B8%20with%20x1%20and%20the%20regression%20coefficients%20in%20the%20regression%20of%20%CE%B8%20on%20any%20pair%20of%20variables%20in%20x1%20and%20x3%20in%20conditions....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=434>) · Original paper and literature note are not hosted.
### P180 - Fig. 6
**Paper:** On the Treatment of Missing Data in Background Questionnaires in Educational Large-Scale Assessments: An Evaluation of Different Procedures
**Original caption:** FIGURE 6. Comparison of the estimated standardized regression coefficients from the regressions of the PVs on any two background variables in the example analysis obtained from different procedures. MAD = mean absolute difference; r = correlation; FCS = fully conditional specification; MIM-PE = missing indicator method with pairwise estimation; MIM-LD = MIM with listwise deletion; MIM-MI = MIM with nested MI.
**Purpose:** Primary method-performance comparison in simulation or the real PISA example. **Visual forms:** Scatter, Composite.
**Physical source pages:** 27.
[PNG](<06_Main_Findings/P180-Fig06-Comparison%20of%20the%20estimated%20standardized%20regression%20coefficients%20from%20the%20regressions%20of%20the%20PVs%20on%20any%20two%20background%20variables%20in%20the%20example%20analysis%20obtained%20from%20different....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=435>) · Original paper and literature note are not hosted.
### P181 - Fig. 3
**Paper:** Participation bias in the UK Biobank distorts genetic associations and downstream analyses
**Original caption:** Fig. 3 | SNP estimates from weighted and unweighted genome-wide analyses. a,b, Summary of the comparison between SNP effects obtained from wGWA and standard GWA analyses on 19 traits. Panel a summarizes the proportions of overestimated and underestimated SNP effects as a result of participation bias. Shown in b are the numbers and proportions of SNPs reaching genome-wide significance in standard GWA, wGWA or both (GWA and wGWA). The scatter plots to the right plot the weighted (|𝛽w|) against the unweighted (|𝛽|) SNP effects for four selected traits.
**Purpose:** Primary effect of participation bias on genetic association estimates. **Visual forms:** Composite.
**Physical source pages:** 5.
[PNG](<06_Main_Findings/P181-Fig03-SNP%20estimates%20from%20weighted%20and%20unweighted%20genome-wide%20analyses.%20a,b,%20Summary%20of%20the%20comparison%20between%20SNP%20effects%20obtained%20from%20wGWA%20and%20standard%20GWA%20analyses%20on%2019%20traits....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=436>) · Original paper and literature note are not hosted.
### P181 - Fig. 4
**Paper:** Participation bias in the UK Biobank distorts genetic associations and downstream analyses
**Original caption:** Fig. 4 | GWA study on the liability to UKBB participation. Shown are the genetic correlations (rg) and corresponding 95% confidence intervals of UKBB participation (n standard GWA = 283,749) with traits indexing participatory behaviour (in green) and other traits (in blue) (including publically available summary statistics generated using standard GWA. SBP, systolic blood pressure; IR,: Item-response theory.
**Purpose:** Primary effect of participation bias on genetic association estimates. **Visual forms:** Composite.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P181-Fig04-GWA%20study%20on%20the%20liability%20to%20UKBB%20participation.%20Shown%20are%20the%20genetic%20correlations%20(rg)%20and%20corresponding%2095%25%20confidence%20intervals%20of%20UKBB%20participation%20(n%20standard%20GWA%20=....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=437>) · Original paper and literature note are not hosted.
### P181 - Fig. 5
**Paper:** Participation bias in the UK Biobank distorts genetic associations and downstream analyses
**Original caption:** Fig. 5 | Weighted SNP heritability and genetic correlation estimates. a, Differences in SNP heritability (h2 DIFF = h2 − h2 w) and genetic correlations (rg,DIFF = |rg| − |rg,w|) obtained from weighted and standard GWA analyses. The diagonal shows the differences in SNP heritability, where biases leading to overestimation (h2 DIFF > 0.02) are plotted in orange and biases leading to underestimation (h2 DIFF < −0.02) are plotted in yellow. The off-diagonal highlights overestimated genetic correlations (rg,DIFF > 0.1) in blue and underestimated genetic correlations (rg,DIFF < −0.1) in green. Tiles coloured in turquoise index genetic correlations where rg and rg,w show opposite directions (with rg printed at the top and rg,w printed at the bottom of the tile). b, Estimates of genetic correlations (rg shown as circles; rg,w shown as triangles) and the corresponding 95% confidence intervals for two selected traits. The asterisks indicate estimates showing significant differences (PFDR < 0.05). All P values are from two-sided tests and are corrected for multiple testing using FDR correction (controlled at 5%).
**Purpose:** Primary effect of participation bias on genetic association estimates. **Visual forms:** Composite.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P181-Fig05-Weighted%20SNP%20heritability%20and%20genetic%20correlation%20estimates.%20a,%20Differences%20in%20SNP%20heritability%20(h2%20DIFF%20=%20h2%20-%20h2%20w)%20and%20genetic%20correlations%20(rg,DIFF%20=%20rg%20-%20rg,w%20)%20obtained....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=438>) · Original paper and literature note are not hosted.
### P181 - Fig. 6
**Paper:** Participation bias in the UK Biobank distorts genetic associations and downstream analyses
**Original caption:** Fig. 6 | Effect of participation bias on MR estimates of exposure–outcome associations. a,b, Summary of results obtained from weighted (α̂ w) and standard (α̂) MR. MR estimates subject to overestimation (|α̂|−|α̂ w|> 0.1) as a result of participation bias are highlighted in violet. MR estimates subject to underestimation (|α̂|−|α̂ w|<−0.1) are highlighted in cyan. The asterisks highlight results where α̂ and α̂ w showed significant (PFDR < 0.05) differences. The error bars (b) indicate the 95% confidence intervals corresponding to α̂ and α̂ w. All P values are from two-sided tests and are corrected for multiple testing using FDR correction (controlled at 5%).
**Purpose:** Primary effect of participation bias on genetic association estimates. **Visual forms:** Composite.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P181-Fig06-Effect%20of%20participation%20bias%20on%20MR%20estimates%20of%20exposure-outcome%20associations.%20a,b,%20Summary%20of%20results%20obtained%20from%20weighted%20(%CE%B1%CC%82%20w)%20and%20standard%20(%CE%B1%CC%82)%20MR.%20MR%20estimates%20subject....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=439>) · Original paper and literature note are not hosted.
### P182 - Fig. 4
**Paper:** Patterns of item nonresponse behaviour to survey questionnaires are systematic and associated with genetic loci
**Original caption:** Fig. 4 | Miami plot for GWAS of the PNA and IDK factors. Miami plot of the resulting P values from a GWAS of the PNA factor (top) and the IDK factor (bottom) for the 360,628 UK Biobank participants included in this study. The x axis is the chromosomal position and the y axis represents the −log10 of the corresponding P value for each location measured in the genome with a two-sided test. The dashed red line marks the threshold for genome-wide significance (P = 5 × 10−8), which corrects for multiple hypothesis testing.
**Purpose:** The Manhattan/Miami display reports the main genome-wide association results for PNA and IDK. **Visual forms:** Composite, Manhattan plot.
**Physical source pages:** 9.
[PNG](<06_Main_Findings/P182-Fig04-Miami%20plot%20for%20GWAS%20of%20the%20PNA%20and%20IDK%20factors.%20Miami%20plot%20of%20the%20resulting%20P%20values%20from%20a%20GWAS%20of%20the%20PNA%20factor%20(top)%20and%20the%20IDK%20factor%20(bottom)%20for%20the%20360,628%20UK%20Biobank....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=440>) · Original paper and literature note are not hosted.
### P183 - Fig. 1
**Paper:** PISA 2018 in England, Northern Ireland, Scotland and Wales: Is the data really representative of all four corners of the UK?
**Original caption:** Figure 1. Trends in PISA mathematics scores across the UK. Source: Sizmur et al. (2019a,b,c).
**Purpose:** Compares UK-nation mathematics scores across PISA cycles. **Visual forms:** Line/profile plot.
**Physical source pages:** 2.
[PNG](<06_Main_Findings/P183-Fig01-Trends%20in%20PISA%20mathematics%20scores%20across%20the%20UK.%20Source%20Sizmur%20et%20al.%20(2019a,b,c).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=441>) · Original paper and literature note are not hosted.
### P183 - Table 3t
**Paper:** PISA 2018 in England, Northern Ireland, Scotland and Wales: Is the data really representative of all four corners of the UK?
**Original caption:** Table 3. Cumulative impact of exclusions, ineligibility, withdrawals and nonresponse across countries.
**Purpose:** Color-coded cross-country or national-versus-sample comparison used in the findings. **Visual forms:** Complex table.
**Physical source pages:** 18.
[PNG](<06_Main_Findings/P183-Table3t-Cumulative%20impact%20of%20exclusions,%20ineligibility,%20withdrawals%20and%20nonresponse%20across%20countries.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=442>) · Original paper and literature note are not hosted.
### P183 - Fig. 5
**Paper:** PISA 2018 in England, Northern Ireland, Scotland and Wales: Is the data really representative of all four corners of the UK?
**Original caption:** Figure 5. The percentage of pupils classified as ‘ineligible/withdrawn’ from the PISA 2015 and 2018 studies.Notes: Dashed red lines refer to the OECD median. Data plotted for OECD countries only. Figures for Scotland and the UK as a whole are illustrated by purple square markers. Source: OECD (2016) and OECD (2019). The Spearman correlation between data points equals 0.65. Notes: Dashed red lines refer to the OECD median. Data plotted for OECD countries only. Figures for Scotland and the UK as a whole are illustrated by purple square markers. Source: OECD (2016) and OECD (2019). The Spearman correlation between data points equals 0.65. 12 of 41 J. Jerrim © 2021 The Authors. Review of Education published by John Wiley & Sons Ltd on behalf of British Educational Research Association
**Purpose:** Compares ineligibility and withdrawal rates between PISA 2015 and 2018. **Visual forms:** Line/profile plot.
**Physical source pages:** 12.
[PNG](<06_Main_Findings/P183-Fig05-The%20percentage%20of%20pupils%20classified%20as%20%E2%80%98ineligible%20withdrawn%E2%80%99%20from%20the%20PISA%202015%20and%202018%20studies.Notes%20Dashed%20red%20lines%20refer%20to%20the%20OECD%20median.%20Data%20plotted%20for%20OECD....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=443>) · Original paper and literature note are not hosted.
### P183 - Table 6t
**Paper:** PISA 2018 in England, Northern Ireland, Scotland and Wales: Is the data really representative of all four corners of the UK?
**Original caption:** Table 6. GCSE mathematics grade distribution. Comparison of the PISA 2018 sample to the national grade distribution.
**Purpose:** Color-coded cross-country or national-versus-sample comparison used in the findings. **Visual forms:** Complex table.
**Physical source pages:** 25.
[PNG](<06_Main_Findings/P183-Table6t-GCSE%20mathematics%20grade%20distribution.%20Comparison%20of%20the%20PISA%202018%20sample%20to%20the%20national%20grade%20distribution.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=444>) · Original paper and literature note are not hosted.
### P183 - Table 7
**Paper:** PISA 2018 in England, Northern Ireland, Scotland and Wales: Is the data really representative of all four corners of the UK?
**Original caption:** Table 7. Prior achievement of responding and nonresponding schools in Northern Ireland Notes: Figures refer to column percentages. Figures may not sum to 100% due to rounding. Small figures have been supressed in places. Figures in panel (a) drawn from NFER (2021b: Tables 2.1–2.3). ‘Nonresponding’ refers to the initially drawn sample of schools that did not get replaced. Participants refers to all participating schools (whether main sample or replacement). Figures in panel (b) drawn from NFER (2021b: Tables 4.1– 4.3). Unweighted number of schools reported in Column 4, whereas the original NRBA presented the weighted number. PISA 2018 across the UK 29 of 41 © 2021 The Authors. Review of Education published by John Wiley & Sons Ltd on behalf of British Educational Research Association
**Purpose:** Research figure visually checked in its article context. **Visual forms:** Complex table.
**Physical source pages:** 29.
[PNG](<06_Main_Findings/P183-Table07-Prior%20achievement%20of%20responding%20and%20nonresponding%20schools%20in%20Northern%20Ireland%20Notes%20Figures%20refer%20to%20column%20percentages.%20Figures%20may%20not%20sum%20to%20100%25%20due%20to%20rounding.%20Small....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=445>) · Original paper and literature note are not hosted.
### P185 - Fig. 1
**Paper:** Retrieval Practice Produces More Learning than Elaborative Studying with Concept Mapping
**Original caption:** Fig. 1. Results of Experiment 1. (A and B) show the proportions correct on verbatim and inference short-answer questions, respectively. (C) shows the proportion of information subjects predicted they would recall on the final test (their metacognitive judgments of learning). Error bars indicate SEM. On the final short-answer test, retrieval practice enhanced long-term learning above and beyond elaborative study with concept mapping by one and a half standard deviations (d = 1.50), yet students were largely unable to predict this benefit.
**Purpose:** Primary experimental comparison of learning and metacognitive predictions. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 3.
[PNG](<06_Main_Findings/P185-Fig01-Results%20of%20Experiment%201.%20(A%20and%20B)%20show%20the%20proportions%20correct%20on%20verbatim%20and%20inference%20short-answer%20questions,%20respectively.%20(C)%20shows%20the%20proportion%20of%20information%20subjects....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=446>) · Original paper and literature note are not hosted.
### P188 - Table 2
**Paper:** Socioeconomic status and student learning: insights from an umbrella review
**Original caption:** Table 2 SES Conceptualization/measurement, relationship between ses and student learning, and mediators of SES
**Purpose:** Cross-review evidence synthesis of SES measurement, learning associations and mediating processes. **Visual forms:** Complex table.
**Physical source pages:** 18, 19, 20, 21, 22, 23, 24, 25, 26, 27.
[PNG part 1](<06_Main_Findings/P188-Table02-SES%20Conceptualization%20measurement,%20relationship%20between%20ses%20and%20student%20learning,%20and%20mediators%20of%20SES-Part01.png>) · [PNG part 2](<06_Main_Findings/P188-Table02-SES%20Conceptualization%20measurement,%20relationship%20between%20ses%20and%20student%20learning,%20and%20mediators%20of%20SES-Part02.png>) · [PNG part 3](<06_Main_Findings/P188-Table02-SES%20Conceptualization%20measurement,%20relationship%20between%20ses%20and%20student%20learning,%20and%20mediators%20of%20SES-Part03.png>) · [PNG part 4](<06_Main_Findings/P188-Table02-SES%20Conceptualization%20measurement,%20relationship%20between%20ses%20and%20student%20learning,%20and%20mediators%20of%20SES-Part04.png>) · [PNG part 5](<06_Main_Findings/P188-Table02-SES%20Conceptualization%20measurement,%20relationship%20between%20ses%20and%20student%20learning,%20and%20mediators%20of%20SES-Part05.png>) · [PNG part 6](<06_Main_Findings/P188-Table02-SES%20Conceptualization%20measurement,%20relationship%20between%20ses%20and%20student%20learning,%20and%20mediators%20of%20SES-Part06.png>) · [PNG part 7](<06_Main_Findings/P188-Table02-SES%20Conceptualization%20measurement,%20relationship%20between%20ses%20and%20student%20learning,%20and%20mediators%20of%20SES-Part07.png>) · [PNG part 8](<06_Main_Findings/P188-Table02-SES%20Conceptualization%20measurement,%20relationship%20between%20ses%20and%20student%20learning,%20and%20mediators%20of%20SES-Part08.png>) · [PNG part 9](<06_Main_Findings/P188-Table02-SES%20Conceptualization%20measurement,%20relationship%20between%20ses%20and%20student%20learning,%20and%20mediators%20of%20SES-Part09.png>) · [PNG part 10](<06_Main_Findings/P188-Table02-SES%20Conceptualization%20measurement,%20relationship%20between%20ses%20and%20student%20learning,%20and%20mediators%20of%20SES-Part10.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=447>) · Original paper and literature note are not hosted.
### P188 - Table 5
**Paper:** Socioeconomic status and student learning: insights from an umbrella review
**Original caption:** Table 5 Contributions of Umbrella review
**Purpose:** Structured synthesis of the review's construct, evidence and boundary-condition contributions. **Visual forms:** Complex table.
**Physical source pages:** 54.
[PNG](<06_Main_Findings/P188-Table05-Contributions%20of%20Umbrella%20review.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=457>) · Original paper and literature note are not hosted.
### P188 - Table 6
**Paper:** Socioeconomic status and student learning: insights from an umbrella review
**Original caption:** Table 6 Nuanced or mixed findings
**Purpose:** Synthesizes mixed evidence across mediation and moderation domains. **Visual forms:** Complex table.
**Physical source pages:** 57.
[PNG](<06_Main_Findings/P188-Table06-Nuanced%20or%20mixed%20findings.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=458>) · Original paper and literature note are not hosted.
### P190 - Fig. 2
**Paper:** Student test-taking effort in low-stakes assessments: evidence from the English version of the PISA 2015 science test
**Original caption:** Fig. 2 Boxplot of student median RT by cluster position (outliers above 200 s are not displayed)
**Purpose:** Shows the response-time distribution across cluster positions and class-enumeration diagnostics. **Visual forms:** Density/histogram/boxplot.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P190-Fig02-Boxplot%20of%20student%20median%20RT%20by%20cluster%20position%20(outliers%20above%20200%20s%20are%20not%20displayed).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=459>) · Original paper and literature note are not hosted.
### P190 - Fig. 3
**Paper:** Student test-taking effort in low-stakes assessments: evidence from the English version of the PISA 2015 science test
**Original caption:** Fig. 3 Elbow plot of the class enumeration for item S131Q02 in cluster position 4
**Purpose:** Shows the response-time distribution across cluster positions and class-enumeration diagnostics. **Visual forms:** Line/profile plot.
**Physical source pages:** 11.
[PNG](<06_Main_Findings/P190-Fig03-Elbow%20plot%20of%20the%20class%20enumeration%20for%20item%20S131Q02%20in%20cluster%20position%204.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=460>) · Original paper and literature note are not hosted.
### P191 - Fig. 2
**Paper:** Substitute or Complement? Generative AI Use and Student Achievement: Evidence from PISA 2025
**Original caption:** Figure 2: AI use and achievement across specifications. Notes: Coefficient on the standardized chatbot index, in PISA points per standard deviation, with 95 percent intervals that combine 80 Fay replicate weights with Rubin imputation variance across ten plausible values. Senate-weighted with within-sample renormalization. The first four rows add controls cumulatively; the last replaces country with school fixed effects. Every row uses the same 249,077 students in 35 countries. Response-style controls are a straight-lining indicator and the within-student standard deviation across the standardized items and its square. Exact estimates and standard errors are in Online Appendix Table A.1.
**Purpose:** Reports the primary achievement association across the paper’s control and fixed-effect specifications. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P191-Fig02-AI%20use%20and%20achievement%20across%20specifications.%20Notes%20Coefficient%20on%20the%20standardized%20chatbot%20index,%20in%20PISA%20points%20per%20standard%20deviation,%20with%2095%20percent%20intervals%20that%20combine....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=461>) · Original paper and literature note are not hosted.
### P191 - Fig. 3
**Paper:** Substitute or Complement? Generative AI Use and Student Achievement: Evidence from PISA 2025
**Original caption:** Figure 3: Average score by group of the AI-use index, after adjustment. Notes: Both the AI-use index and the score are adjusted for country fixed effects, socioeconomic status and the response-style terms, that is, replaced by their difference from what those variables predict (plus the mean). Students are sorted by the adjusted index into 20 groups of about 5 percent; each point is a group’s average. The line’s slope is the coefficient in the fourth row of Figure 2, with its 95 percent interval shaded. Higher index values mean more AI use.
**Purpose:** Shows the adjusted achievement gradient that corresponds to the principal model estimate. **Visual forms:** Composite, Line/profile plot.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P191-Fig03-Average%20score%20by%20group%20of%20the%20AI-use%20index,%20after%20adjustment.%20Notes%20Both%20the%20AI-use%20index%20and%20the%20score%20are%20adjusted%20for%20country%20fixed%20effects,%20socioeconomic%20status%20and%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=462>) · Original paper and literature note are not hosted.
### P191 - Fig. 4
**Paper:** Substitute or Complement? Generative AI Use and Student Achievement: Evidence from PISA 2025
**Original caption:** Figure 4: The association by kind of AI use. Notes: PISA points per standard deviation of the item or component (one standard deviation of an item is about 1.4 response categories), with 95 percent intervals, for 249,077 students in 35 OECD countries. The top four rows enter the four chatbot items together, so each is read holding the other three fixed (variance inflation factors 2.0 to 2.5); the bottom two enter the first two principal components together, which are uncorrelated by construction. Filled markers condition on country fixed effects and hollow markers on school fixed effects; both include socioeconomic status and the response-style terms. Exact estimates and standard errors are in Online Appendix Table A.2.
**Purpose:** Compares estimated achievement associations for different AI-use purposes and level/purpose components. **Visual forms:** Composite, Forest/coefficient plot.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P191-Fig04-The%20association%20by%20kind%20of%20AI%20use.%20Notes%20PISA%20points%20per%20standard%20deviation%20of%20the%20item%20or%20component%20(one%20standard%20deviation%20of%20an%20item%20is%20about%201.4%20response%20categories),%20with....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=463>) · Original paper and literature note are not hosted.
### P196 - Fig. 4
**Paper:** Universal Basic Incomes vs. Targeted Transfers: Anti-Poverty Programs in Developing Countries
**Original caption:** Figure 4: Tradeoffs between inclusion error and exclusion error by varying eligibility cutoff Panel A: ROC Curves for Program Targeting Panel B: Benefits per Household vs. Inclusion Error for Different Eligibility Cutoffs 𝑐 Notes: Panel A plots the ROC curves for both Peru and Indonesia, using a target consumption value equal to the government poverty line in Peru and 1.5x the government poverty line in Indonesia. Panel B traces the per-household benefit amount as inclusion error increases, assuming a fixed transfer budget of approximately $1.83 billion per year in Indonesia and $274 million per year in Peru (modeled after the sizes of BLT and Juntos, respectively). We include cost-savings from administrative targeting at the universal basic income point (inclusion error = 1).
**Purpose:** Primary targeting inclusion/exclusion error and transfer-benefit trade-off. **Visual forms:** Composite, Line plot.
**Physical source pages:** 32.
[PNG](<06_Main_Findings/P196-Fig04-Tradeoffs%20between%20inclusion%20error%20and%20exclusion%20error%20by%20varying%20eligibility%20cutoff%20Panel%20A%20ROC%20Curves%20for%20Program%20Targeting%20Panel%20B%20Benefits%20per%20Household%20vs.%20Inclusion%20Error....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=464>) · Original paper and literature note are not hosted.
### P196 - Fig. 5
**Paper:** Universal Basic Incomes vs. Targeted Transfers: Anti-Poverty Programs in Developing Countries
**Original caption:** Figure 5: Social Welfare vs. Inclusion Error Notes: Figure 5 uses the same pooled household data as in Figure 4. For each cutoff value c, we calculate the percapita benefit amount for included households, and then calculate constant relative risk-aversion utility using the formula noted above. If a household is not included in the program at a given value of c, we simply set bi = 0. Dashed lines indicate the point of maximum social welfare in each country.
**Purpose:** Primary welfare comparison along targeting trade-offs. **Visual forms:** Line plot.
**Physical source pages:** 33.
[PNG](<06_Main_Findings/P196-Fig05-Social%20Welfare%20vs.%20Inclusion%20Error%20Notes%20Figure%205%20uses%20the%20same%20pooled%20household%20data%20as%20in%20Figure%204.%20For%20each%20cutoff%20value%20c,%20we%20calculate%20the%20percapita%20benefit%20amount%20for....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=465>) · Original paper and literature note are not hosted.
### P196 - Fig. 6
**Paper:** Universal Basic Incomes vs. Targeted Transfers: Anti-Poverty Programs in Developing Countries
**Original caption:** Figure 6: Horizontal Equity vs. Inclusion Error Notes: Figure 6 traces horizontal equity as a function of inclusion error. At each cutoff c, we calculate, for each household, the percentage of households within ±5 income percentiles (based on actual income) that received the same benefit status – included or excluded – based on the results of proxy-means test prediction. In other words, for households that were included in the program at a given c, we calculate the percentage of similar households that were also included; for households that were excluded, we calculate the percentage of similar households that were also excluded. At each cutoff point, we average these percentages over all households to compute the horizontal equity variable graphed above. Dashed lines indicate the location and value of minimum horizontal equity.
**Purpose:** Primary horizontal equity comparison along targeting trade-offs. **Visual forms:** Line plot.
**Physical source pages:** 34.
[PNG](<06_Main_Findings/P196-Fig06-Horizontal%20Equity%20vs.%20Inclusion%20Error%20Notes%20Figure%206%20traces%20horizontal%20equity%20as%20a%20function%20of%20inclusion%20error.%20At%20each%20cutoff%20c,%20we%20calculate,%20for%20each%20household,%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=466>) · Original paper and literature note are not hosted.
### P197 - Fig. 3
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** FIGURE 3 Plot of the 1600 θ̂ i (left panels) and SÊ(θ̂)i (right panels) by data-generating mechanisms, for the three analysis methods. The vertical axis is repetition number, to provide some separation between points. The yellow pipes are sample means [Colour figure can be viewed at wileyonlinelibrary.com]
**Purpose:** Displays estimates and standard errors from simulation repetitions. **Visual forms:** Composite, Scatter.
**Physical source pages:** 19.
[PNG](<06_Main_Findings/P197-Fig03-Plot%20of%20the%201600%20%CE%B8%CC%82%20i%20(left%20panels)%20and%20S%C3%8A(%CE%B8%CC%82)i%20(right%20panels)%20by%20data-generating%20mechanisms,%20for%20the%20three%20analysis%20methods.%20The%20vertical%20axis%20is%20repetition%20number,%20to%20provide....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=467>) · Original paper and literature note are not hosted.
### P197 - Fig. 4
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** FIGURE 4 Comparison of estimates for methods when γ = 1.5, where each point represents one repetition. A, Upper triangle displays θ̂ i; lower triangle displays SÊ(θ̂ i); B, Plot of difference vs mean for θ̂i and SÊ(θ̂i), with Weibull as the comparator
**Purpose:** Compares paired estimates and standard errors across methods. **Visual forms:** Composite, Scatter.
**Physical source pages:** 20.
[PNG](<06_Main_Findings/P197-Fig04-Comparison%20of%20estimates%20for%20methods%20when%20%CE%B3%20=%201.5,%20where%20each%20point%20represents%20one%20repetition.%20A,%20Upper%20triangle%20displays%20%CE%B8%CC%82%20i;%20lower%20triangle%20displays%20S%C3%8A(%CE%B8%CC%82%20i);%20B,%20Plot%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=468>) · Original paper and literature note are not hosted.
### P197 - Fig. 5
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** FIGURE 5 “Zip plot” of the 1600 confidence intervals for each data-generating mechanism and analysis method. The vertical axis is the fractional centile of |z| with z =(𝜃̂ i−𝜃)∕ModSE associated with the confidence interval [Colour figure can be viewed at wileyonlinelibrary.com]
**Purpose:** Visualizes confidence-interval coverage across simulation mechanisms and methods. **Visual forms:** Composite, Zip plot.
**Physical source pages:** 20.
[PNG](<06_Main_Findings/P197-Fig05-%E2%80%9CZip%20plot%E2%80%9D%20of%20the%201600%20confidence%20intervals%20for%20each%20data-generating%20mechanism%20and%20analysis%20method.%20The%20vertical%20axis%20is%20the%20fractional%20centile%20of%20z%20with%20z%20=(%CE%B8%CC%82%20i-%CE%B8)%E2%88%95ModSE....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=469>) · Original paper and literature note are not hosted.
### P197 - Fig. 6
**Paper:** Using simulation studies to evaluate statistical methods
**Original caption:** FIGURE 6 Lollipop plot of performance for measures of interest (Monte Carlo 95% confidence intervals in parentheses). Concerning features need not be highlighted since they are readily visible. See, also, Table 8
**Purpose:** Summarizes comparative performance and Monte Carlo uncertainty. **Visual forms:** Composite, Lollipop plot.
**Physical source pages:** 21.
[PNG](<06_Main_Findings/P197-Fig06-Lollipop%20plot%20of%20performance%20for%20measures%20of%20interest%20(Monte%20Carlo%2095%25%20confidence%20intervals%20in%20parentheses).%20Concerning%20features%20need%20not%20be%20highlighted%20since%20they%20are%20readily....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=470>) · Original paper and literature note are not hosted.
### P201 - Fig. 4
**Paper:** A psychometrics of individual differences in experimental tasks
**Original caption:** Fig. 4 The lack of correlation between flanker task and Stroop task performance. Left: Scatter plot of individuals’ model-based estimates. Right: Posterior distribution of the correlation coefficient
**Purpose:** Main substantive finding of little latent Stroop–flanker correlation. **Visual forms:** Scatter, Density/histogram/boxplot, Composite.
**Physical source pages:** 9.
[PNG](<06_Main_Findings/P201-Fig04-The%20lack%20of%20correlation%20between%20flanker%20task%20and%20Stroop%20task%20performance.%20Left%20Scatter%20plot%20of%20individuals%E2%80%99%20model-based%20estimates.%20Right%20Posterior%20distribution%20of%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=878>) · Original paper and literature note are not hosted.
### P203 - Fig. 2
**Paper:** A Systematic Review and Meta-Analysis of Discrepancies Between Logged and Self-Reported Digital Media Use
**Original caption:** Figure 2. Forest plot of the effect sizes for studies included in the meta-analysis for the association between self-reported and logged measures of digital media use. Individual Pearson’s r estimates are depicted by filled squares, with the square sizes indicating the relative weight of each effect size estimate in the meta-analysis. The filled diamond represents the overall summary effect size (r = 0.38, 95% CI [0.33, 0.42], p < 0.001). The error bars and diamond width represent the 95% CIs for the effect sizes. The dashed reference line at the intercept for r = 0.5 represents the point from which the magnitude of the association would be sufficient to conclude that the measures are appropriate substitutes for one another. RE = Random effects model. RVE = Robust variance estimation (conducted with a correlated effects weighting scheme).
**Purpose:** Pooled effects quantify agreement or reporting accuracy between measurement sources. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 26, 39.
[PNG](<06_Main_Findings/P203-Fig02-Forest%20plot%20of%20the%20effect%20sizes%20for%20studies%20included%20in%20the%20meta-analysis%20for%20the%20association%20between%20self-reported%20and%20logged%20measures%20of%20digital%20media%20use.%20Individual....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=879>) · Original paper and literature note are not hosted.
### P203 - Fig. 4
**Paper:** A Systematic Review and Meta-Analysis of Discrepancies Between Logged and Self-Reported Digital Media Use
**Original caption:** Figure 4. Forest plot of the effect sizes for studies included in the meta-analysis for the association between self-reported problematic use and logged measures of use. Individual Pearson’s r estimates are depicted by filled squares, with the square sizes indicating the relative weight of each effect size estimate in the meta-analysis. The filled diamond represents the overall summary effect size (r = 0.25, 95% CI [0.20, 0.29], p < 0.001). The error bars and diamond width represent the 95% CIs for the effect sizes. The dashed reference line at the intercept for r = 0.5 represents the point from which the magnitude of the association would be sufficient to conclude that the measures are appropriate substitutes for one another. RE = Random effects model. RVE = Robust variance estimation (conducted with a correlated effects weighting scheme).
**Purpose:** Pooled effects quantify agreement or reporting accuracy between measurement sources. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 26, 41.
[PNG](<06_Main_Findings/P203-Fig04-Forest%20plot%20of%20the%20effect%20sizes%20for%20studies%20included%20in%20the%20meta-analysis%20for%20the%20association%20between%20self-reported%20problematic%20use%20and%20logged%20measures%20of%20use.%20Individual....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=880>) · Original paper and literature note are not hosted.
### P203 - Fig. 5
**Paper:** A Systematic Review and Meta-Analysis of Discrepancies Between Logged and Self-Reported Digital Media Use
**Original caption:** Figure 5. Forest plot of the effect sizes for studies included in the meta-analysis for the ratio of means between self-reported and logged measures of digital media use. The results are represented on a log scale. Individual response ratios (ratio of means) are depicted by filled squares, with the square sizes indicating the relative weight of each effect size estimate in the meta-analysis. The filled diamond represents the overall summary effect size (R = 1.21, 95% CI [0.94, 1.54], p = 0.129). The error bars and diamond width represent the 95% CIs for the effect sizes. The dashed reference line at the intercept for 1.0 represents a 1:1 ratio between self760 reported and logged digital media use, with values below one indicating under-reporting and values above one indicating over-reporting of digital media use. RE = Random effects model. RVE = Robust variance estimation (conducted with a correlated effects weighting scheme).
**Purpose:** Pooled effects quantify agreement or reporting accuracy between measurement sources. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 27, 42.
[PNG](<06_Main_Findings/P203-Fig05-Forest%20plot%20of%20the%20effect%20sizes%20for%20studies%20included%20in%20the%20meta-analysis%20for%20the%20ratio%20of%20means%20between%20self-reported%20and%20logged%20measures%20of%20digital%20media%20use.%20The%20results%20are....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=881>) · Original paper and literature note are not hosted.
### P205 - Fig. 8
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** Figure 8 Essay score improvements of four groups after revision
**Purpose:** Primary essay-improvement distributions by randomized condition, reported in appendix. **Visual forms:** Boxplot, Scatter overlay.
**Physical source pages:** 42.
[PNG](<06_Main_Findings/P205-Fig08-Essay%20score%20improvements%20of%20four%20groups%20after%20revision.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=882>) · Original paper and literature note are not hosted.
### P205 - Fig. 9
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** Figure 9 Essay scores of four groups after revision
**Purpose:** Postrevision essay-score distributions by randomized support condition. **Visual forms:** Boxplot, Scatter overlay.
**Physical source pages:** 43.
[PNG](<06_Main_Findings/P205-Fig09-Essay%20scores%20of%20four%20groups%20after%20revision.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=883>) · Original paper and literature note are not hosted.
### P206 - Table 1
**Paper:** Beyond traditional interviews: Psychometric analysis of asynchronous video interviews for personality and interview performance evaluation using machine learning
**Original caption:** Table 1 Summary of hypotheses, open research questions, and findings of the present study. Notes. AVI = asynchronous video interview; LIWC = Linguistic Inquiry and Word Count (Pennebaker et al., 2015).
**Purpose:** Structured hypothesis/question-to-finding evidence map summarizes the study. **Visual forms:** Complex table.
**Physical source pages:** 5.
[PNG](<06_Main_Findings/P206-Table01-Summary%20of%20hypotheses,%20open%20research%20questions,%20and%20findings%20of%20the%20present%20study.%20Notes.%20AVI%20=%20asynchronous%20video%20interview;%20LIWC%20=%20Linguistic%20Inquiry%20and%20Word%20Count....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=884>) · Original paper and literature note are not hosted.
### P207 - Fig. 1
**Paper:** ChatGPT improves creative problem-solving performance in university students: An experimental study
**Original caption:** Fig. 1. Originality of solutions in pre-test and experimental task Note. The pre-test task was an Unusual Uses Task (solved without ChatGPT in both conditions); the experimental task was a Product Improvement Task (solved with ChatGPT in the experimental group and without ChatGPT in the control group).
**Purpose:** Primary originality comparison between randomized groups across pretest and experimental task; retain full note. **Visual forms:** Line plot, Error bars.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P207-Fig01-Originality%20of%20solutions%20in%20pre-test%20and%20experimental%20task%20Note.%20The%20pre-test%20task%20was%20an%20Unusual%20Uses%20Task%20(solved%20without%20ChatGPT%20in%20both%20conditions);%20the%20experimental%20task....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=885>) · Original paper and literature note are not hosted.
### P207 - Fig. 2
**Paper:** ChatGPT improves creative problem-solving performance in university students: An experimental study
**Original caption:** Fig. 2. Self-efficacy to resolve the problem-solving task without and with ChatGPT.
**Purpose:** Primary self-efficacy comparison; corrected lower crop to retain 95% CI key and true caption. **Visual forms:** Bar chart, Error bars.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P207-Fig02-Self-efficacy%20to%20resolve%20the%20problem-solving%20task%20without%20and%20with%20ChatGPT.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=886>) · Original paper and literature note are not hosted.
### P207 - Fig. 3
**Paper:** ChatGPT improves creative problem-solving performance in university students: An experimental study
**Original caption:** Fig. 3. Perceived task interest, difficulty, and mental effort invested in task resolution
**Purpose:** Primary task-interest, difficulty and effort comparison. **Visual forms:** Bar chart, Error bars.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P207-Fig03-Perceived%20task%20interest,%20difficulty,%20and%20mental%20effort%20invested%20in%20task%20resolution.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=887>) · Original paper and literature note are not hosted.
### P213 - Fig. 4
**Paper:** COVID-19 school closures, learning losses and intergenerational mobility
**Original caption:** Fig. 4 Evolution of educational attainment over time and based on the COVID-19 scenario. Authors’ estimates based on data from Azevedo et al. (2022) and GDIM. Estimates based on the years of education of the 1980s cohort, and thus rely on the assumption that the years of education of the 1980s cohort roughly reflect the years of education for the current school-going cohort (roughly 2000–2015). Estimates in Fig. 1 suggest that this is not a very strong assumption.
**Purpose:** Primary simulated educational-attainment trajectory with and without pandemic losses. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P213-Fig04-Evolution%20of%20educational%20attainment%20over%20time%20and%20based%20on%20the%20COVID-19%20scenario.%20Authors%E2%80%99%20estimates%20based%20on%20data%20from%20Azevedo%20et%20al.%20(2022)%20and%20GDIM.%20Estimates%20based%20on%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=888>) · Original paper and literature note are not hosted.
### P213 - Fig. 5
**Paper:** COVID-19 school closures, learning losses and intergenerational mobility
**Original caption:** Fig. 5 Evolution of absolute mobility by birth cohort. Authors’ estimates based on data from Azevedo et al. (2022) and GDIM
**Purpose:** Primary projected absolute educational-mobility comparison against historical cohorts. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P213-Fig05-Evolution%20of%20absolute%20mobility%20by%20birth%20cohort.%20Authors%E2%80%99%20estimates%20based%20on%20data%20from%20Azevedo%20et%20al.%20(2022)%20and%20GDIM.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=889>) · Original paper and literature note are not hosted.
### P215 - Fig. 2
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Fig. 2. Key findings by research question.
**Purpose:** Synthesis of the three research questions and their results. **Visual forms:** Summary diagram.
**Physical source pages:** 16.
[PNG](<06_Main_Findings/P215-Fig02-Key%20findings%20by%20research%20question.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=890>) · Original paper and literature note are not hosted.
### P215 - Fig. B1
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Appendix B1. Forest plots of effect sizes for academic performance.
**Purpose:** Original pooled-outcome results with study statistics, confidence intervals and summary diamond; all columns and caption retained. **Visual forms:** Forest plot.
**Physical source pages:** 25.
[PNG](<06_Main_Findings/P215-FigB1-Appendix%20B1.%20Forest%20plots%20of%20effect%20sizes%20for%20academic%20performance.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=891>) · Original paper and literature note are not hosted.
### P215 - Fig. B2
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Appendix B2. Forest plots of effect sizes for affective-motivational states.
**Purpose:** Original pooled-outcome results with study statistics, confidence intervals and summary diamond; all columns and caption retained. **Visual forms:** Forest plot.
**Physical source pages:** 26.
[PNG](<06_Main_Findings/P215-FigB2-Appendix%20B2.%20Forest%20plots%20of%20effect%20sizes%20for%20affective-motivational%20states.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=892>) · Original paper and literature note are not hosted.
### P215 - Fig. B3
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Appendix B3. Forest plots of effect sizes for higher-order thinking propensities.
**Purpose:** Original pooled-outcome results with study statistics, confidence intervals and summary diamond; all columns and caption retained. **Visual forms:** Forest plot.
**Physical source pages:** 26.
[PNG](<06_Main_Findings/P215-FigB3-Appendix%20B3.%20Forest%20plots%20of%20effect%20sizes%20for%20higher-order%20thinking%20propensities.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=893>) · Original paper and literature note are not hosted.
### P215 - Fig. B4
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Appendix B4. Forest plots of effect sizes for self-efficacy.
**Purpose:** Original pooled-outcome results with study statistics, confidence intervals and summary diamond; all columns and caption retained. **Visual forms:** Forest plot.
**Physical source pages:** 26.
[PNG](<06_Main_Findings/P215-FigB4-Appendix%20B4.%20Forest%20plots%20of%20effect%20sizes%20for%20self-efficacy.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=894>) · Original paper and literature note are not hosted.
### P215 - Fig. B5
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Appendix B5. Forest plots of effect sizes for mental effort.
**Purpose:** Original pooled-outcome results with study statistics, confidence intervals and summary diamond; all columns and caption retained. **Visual forms:** Forest plot.
**Physical source pages:** 27.
[PNG](<06_Main_Findings/P215-FigB5-Appendix%20B5.%20Forest%20plots%20of%20effect%20sizes%20for%20mental%20effort.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=895>) · Original paper and literature note are not hosted.
### P216 - Table 4
**Paper:** Does geographical location impact educational disparities among Ecuadorians? A novel two-stage inequality decomposition method
**Original caption:** Table 4. Ecuadorian educational inequality and its components (2014 and 2021).
**Purpose:** Nonstandard spatial/within-component decomposition layout, arranged as an explicit hierarchy rather than an ordinary regression table. **Visual forms:** Hierarchical decomposition, Complex table.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P216-Table04-Ecuadorian%20educational%20inequality%20and%20its%20components%20(2014%20and%202021).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=896>) · Original paper and literature note are not hosted.
### P223 - Fig. 3
**Paper:** Generating group-level scores under response accuracy-time conditional dependence
**Original caption:** Fig. 3 Residual variance estimates in PISA data across IRT-LRMs (M3, M4, M5)
**Purpose:** Displays precision gain and residual-dependence impact across the three models. **Visual forms:** Line/profile plot.
**Physical source pages:** 14.
[PNG](<06_Main_Findings/P223-Fig03-Residual%20variance%20estimates%20in%20PISA%20data%20across%20IRT-LRMs%20(M3,%20M4,%20M5).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=897>) · Original paper and literature note are not hosted.
### P230 - Fig. 3
**Paper:** Impact of AI assistance on student agency
**Original caption:** Fig. 3. Comparing the control group (AI) with the three experimental groups (NR, SR, and SAI) using six measures.
**Purpose:** Primary second-four-week randomized condition comparisons across six measures; all panels retained. **Visual forms:** Violin plot, Boxplot, Composite.
**Physical source pages:** 8.
[PNG](<06_Main_Findings/P230-Fig03-Comparing%20the%20control%20group%20(AI)%20with%20the%20three%20experimental%20groups%20(NR,%20SR,%20and%20SAI)%20using%20six%20measures.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=898>) · Original paper and literature note are not hosted.
### P231 - Fig. 4
**Paper:** Inequality of opportunity in non-cognitive skills: a cross-region analysis
**Original caption:** Fig. 4 Inequality of outcome and inequality of opportunity.
**Purpose:** Primary cross-site relationship between observed inequality of outcomes and estimated inequality of opportunity. **Visual forms:** Scatter plot.
**Physical source pages:** 9.
[PNG](<06_Main_Findings/P231-Fig04-Inequality%20of%20outcome%20and%20inequality%20of%20opportunity.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=899>) · Original paper and literature note are not hosted.
### P235 - Fig. 2
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** Figure 2: Average test score by group and treatment: U.S. vs. Shanghai. Notes: Average score for students who received no incentives (Control) and for students who received incentives (Treatment) by school and track.
**Purpose:** Primary experimental score comparison across treatment, country and school/track. **Visual forms:** Bar chart.
**Physical source pages:** 28.
[PNG](<06_Main_Findings/P235-Fig02-Average%20test%20score%20by%20group%20and%20treatment%20U.S.%20vs.%20Shanghai.%20Notes%20Average%20score%20for%20students%20who%20received%20no%20incentives%20(Control)%20and%20for%20students%20who%20received%20incentives....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=900>) · Original paper and literature note are not hosted.
### P235 - Fig. 3
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** Figure 3: Distribution of test scores by treatment group. Notes: Test of equality of distributions p-values: U.S. p = 0.0031, Shanghai p = 0.36.
**Purpose:** Experimental distributional effect of incentives in the United States and Shanghai. **Visual forms:** Line/profile plot.
**Physical source pages:** 29.
[PNG](<06_Main_Findings/P235-Fig03-Distribution%20of%20test%20scores%20by%20treatment%20group.%20Notes%20Test%20of%20equality%20of%20distributions%20p-values%20U.S.%20p%20=%200.0031,%20Shanghai%20p%20=%200.36.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=901>) · Original paper and literature note are not hosted.
### P235 - Fig. 6
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** Figure 6: Estimated treatment effects on PISA by question order, U.S. Notes: Lines are smoothed using local polynomial regression.
**Purpose:** Transported counterfactual item-position performance under estimated incentive effects. **Visual forms:** Line/profile plot.
**Physical source pages:** 32.
[PNG](<06_Main_Findings/P235-Fig06-Estimated%20treatment%20effects%20on%20PISA%20by%20question%20order,%20U.S.%20Notes%20Lines%20are%20smoothed%20using%20local%20polynomial%20regression.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=902>) · Original paper and literature note are not hosted.
### P235 - Fig. 7
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** Figure 7: Distribution of estimated PISA scores. Notes: The vertical lines in Panel A indicate the mean for Control (442.29) and Treatment (464.37). The vertical lines in Panel B indicate the mean for Control (465.56) and Treatment (489.26). In Panel B, we weight estimated scores in Control to match the estimated U.S. national distribution. We weight estimated scores in Treatment using Control weights for equivalent ranks.
**Purpose:** Main modeled national-score implications, shown before and after national reweighting. **Visual forms:** Density/histogram/boxplot, Composite.
**Physical source pages:** 33.
[PNG](<06_Main_Findings/P235-Fig07-Distribution%20of%20estimated%20PISA%20scores.%20Notes%20The%20vertical%20lines%20in%20Panel%20A%20indicate%20the%20mean%20for%20Control%20(442.29)%20and%20Treatment%20(464.37).%20The%20vertical%20lines%20in%20Panel%20B%20indicate....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=903>) · Original paper and literature note are not hosted.
### P238 - Fig. 5
**Paper:** Profiling low-proficiency science students in the Philippines using machine learning
**Original caption:** Fig. 5 Summary of grid search accuracies of the different machine learning models. The scatterplot illustrating the range of test accuracies during the cross-validation on best machine learning models shows that the RF returned the best accuracies.
**Purpose:** Comparison of predictive accuracy across tuned models. **Visual forms:** Scatter plot.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P238-Fig05-Summary%20of%20grid%20search%20accuracies%20of%20the%20different%20machine%20learning%20models.%20The%20scatterplot%20illustrating%20the%20range%20of%20test%20accuracies%20during%20the%20cross-validation%20on%20best%20machine....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=904>) · Original paper and literature note are not hosted.
### P238 - Fig. 6
**Paper:** Profiling low-proficiency science students in the Philippines using machine learning
**Original caption:** Fig. 6 The top 15 variables that are important in predicting the Science proficiency (poor or better) of Filipino students in PISA 2018, listed in decreasing importance (SHAP value) from top to bottom. Blue bars represent variables that negatively affect the prediction of poor-performing students while red bars indicate that a variable positively affects the prediction of poor-performing students.
**Purpose:** SHAP importance of predictors for the low-proficiency prediction task. **Visual forms:** Bar chart, Feature importance.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P238-Fig06-The%20top%2015%20variables%20that%20are%20important%20in%20predicting%20the%20Science%20proficiency%20(poor%20or%20better)%20of%20Filipino%20students%20in%20PISA%202018,%20listed%20in%20decreasing%20importance%20(SHAP%20value)....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=905>) · Original paper and literature note are not hosted.
### P245 - Fig. 1
**Paper:** Students’ subjective well-being, school bullying, and belonging during the COVID-19 pandemic: Comparison between PISA 2018 and PISA 2022
**Original caption:** Fig. 1 Performance of subjective well-being among students in 2018 and 2022. Students' subjective well-being is divided into four levels, with blue color indicating "Not satisfied (< 5)", orange color indicating “Relatively satisfied (5–6)”, and yellow color indicating “Moderately satisfied (7–8)” and the green color means "Highly satisfied (9–10).
**Purpose:** Primary comparison of subjective well-being categories in the 2018 and 2022 surveys. **Visual forms:** Stacked bar chart.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P245-Fig01-Performance%20of%20subjective%20well-being%20among%20students%20in%202018%20and%202022.%20Students'%20subjective%20well-being%20is%20divided%20into%20four%20levels,%20with%20blue%20color%20indicating%20Not%20satisfied%20(....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=906>) · Original paper and literature note are not hosted.
### P248 - Fig. 1
**Paper:** The effect of funding for disadvantaged students on academic underachievement
**Original caption:** Fig. 1 The relationship between math test scores and the share of disadvantaged students. The figure shows two regression lines of the log of mathematics test score on the share of disadvantaged students in a school—one on each side of the threshold of 10%—while allowing for different slopes.
**Purpose:** Primary visual relationship of mathematics scores to the funding eligibility running variable. **Visual forms:** Scatter plot, Regression discontinuity plot.
**Physical source pages:** 6.
[PNG](<06_Main_Findings/P248-Fig01-The%20relationship%20between%20math%20test%20scores%20and%20the%20share%20of%20disadvantaged%20students.%20The%20figure%20shows%20two%20regression%20lines%20of%20the%20log%20of%20mathematics%20test%20score%20on%20the%20share%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=907>) · Original paper and literature note are not hosted.
### P248 - Fig. 2
**Paper:** The effect of funding for disadvantaged students on academic underachievement
**Original caption:** Fig. 2 Underachievement decomposed into a long term (persistent) and a short term (transient) component. The median is represented by the thick horizontal line.
**Purpose:** Primary decomposition of persistent, transient and overall underachievement. **Visual forms:** Boxplot.
**Physical source pages:** 7.
[PNG](<06_Main_Findings/P248-Fig02-Underachievement%20decomposed%20into%20a%20long%20term%20(persistent)%20and%20a%20short%20term%20(transient)%20component.%20The%20median%20is%20represented%20by%20the%20thick%20horizontal%20line.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=908>) · Original paper and literature note are not hosted.
### P251 - Fig. 3
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 3. Study 1: Correlations between reasoning and higher-order factor scores as well as FFM dimensions scores of personality. Rea = reasoning, ver = verbal, num = numerical, fig = figural.
**Purpose:** Reports personality–intelligence associations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 10.
[PNG](<06_Main_Findings/P251-Fig03-Study%201%20Correlations%20between%20reasoning%20and%20higher-order%20factor%20scores%20as%20well%20as%20FFM%20dimensions%20scores%20of%20personality.%20Rea%20=%20reasoning,%20ver%20=%20verbal,%20num%20=%20numerical,%20fig%20=%20figural.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=909>) · Original paper and literature note are not hosted.
### P251 - Fig. 4
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 4. Study 1: Correlations between reasoning and the facets of (a) Openness, (b) Conscientiousness and (c) Extraversion. Rea = reasoning, ver = verbal, num = numerical, fig = figural.
**Purpose:** Reports personality–intelligence associations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 11.
[PNG](<06_Main_Findings/P251-Fig04-Study%201%20Correlations%20between%20reasoning%20and%20the%20facets%20of%20(a)%20Openness,%20(b)%20Conscientiousness%20and%20(c)%20Extraversion.%20Rea%20=%20reasoning,%20ver%20=%20verbal,%20num%20=%20numerical,%20fig%20=%20figural.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=910>) · Original paper and literature note are not hosted.
### P251 - Fig. 5
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 5. Study 1: Correlations between reasoning and the facets of (a) Agreeableness and (b) Emotional Stability. Rea = reasoning, ver = verbal, num = numerical, fig = figural.
**Purpose:** Reports personality–intelligence associations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 12.
[PNG](<06_Main_Findings/P251-Fig05-Study%201%20Correlations%20between%20reasoning%20and%20the%20facets%20of%20(a)%20Agreeableness%20and%20(b)%20Emotional%20Stability.%20Rea%20=%20reasoning,%20ver%20=%20verbal,%20num%20=%20numerical,%20fig%20=%20figural.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=911>) · Original paper and literature note are not hosted.
### P251 - Fig. 10
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 10. Study 2: Correlations between intelligence and higher-order factors of personality: (a) GFP, (b) Stability, (c) Plasticity. Rea = reasoning, Flu = fluency, Spe = perceptual speed, Mem = memory, Ver = verbal (intelligence), Num = numerical (intelligence), Fig = figural (intelligence).
**Purpose:** Reports personality–intelligence associations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 20.
[PNG](<06_Main_Findings/P251-Fig10-Study%202%20Correlations%20between%20intelligence%20and%20higher-order%20factors%20of%20personality%20(a)%20GFP,%20(b)%20Stability,%20(c)%20Plasticity.%20Rea%20=%20reasoning,%20Flu%20=%20fluency,%20Spe%20=%20perceptual%20speed,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=912>) · Original paper and literature note are not hosted.
### P251 - Fig. 11
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 11. Study 2: Correlations between intelligence and (a) Openness, (b) Conscientiousness and (c) Extraversion. Rea = reasoning, Flu = fluency, Spe = perceptual speed, Mem = memory, Ver = verbal (intelligence), Num = numerical (intelligence), Fig = figural (intelligence).
**Purpose:** Reports personality–intelligence associations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 21.
[PNG](<06_Main_Findings/P251-Fig11-Study%202%20Correlations%20between%20intelligence%20and%20(a)%20Openness,%20(b)%20Conscientiousness%20and%20(c)%20Extraversion.%20Rea%20=%20reasoning,%20Flu%20=%20fluency,%20Spe%20=%20perceptual%20speed,%20Mem%20=%20memory,%20Ver....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=913>) · Original paper and literature note are not hosted.
### P251 - Fig. 12
**Paper:** The Relation of Personality and Intelligence—What Can the Brunswik Symmetry Principle Tell Us?
**Original caption:** Figure 12. Study 2: Correlations between intelligence and (a) Agreeableness and (b) Emotional Stability. Rea = reasoning, Flu = fluency, Spe = perceptual speed, Mem = memory, Ver = verbal (intelligence), Num = numerical (intelligence), Fig = figural (intelligence).
**Purpose:** Reports personality–intelligence associations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 22.
[PNG](<06_Main_Findings/P251-Fig12-Study%202%20Correlations%20between%20intelligence%20and%20(a)%20Agreeableness%20and%20(b)%20Emotional%20Stability.%20Rea%20=%20reasoning,%20Flu%20=%20fluency,%20Spe%20=%20perceptual%20speed,%20Mem%20=%20memory,%20Ver%20=%20verbal....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=914>) · Original paper and literature note are not hosted.
## 07 Mechanisms and Heterogeneity
### P057 - Fig. 1
**Paper:** A social network analysis of student-perceived teaching practices in top-performing countries and regions in reading literacy: Insights from PISA 2018
**Original caption:** Fig. 1. Hierarchical clustering analysis of teaching practices across high-performing countries or regions.
**Purpose:** Clusters national teaching-practice profiles to identify between-system heterogeneity. **Visual forms:** Dendrogram.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P057-Fig01-Hierarchical%20clustering%20analysis%20of%20teaching%20practices%20across%20high-performing%20countries%20or%20regions.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=471>) · Original paper and literature note are not hosted.
### P058 - Fig. 4
**Paper:** Can inquiry-based pedagogy improve math performance? Evidence from 5711 students in Vietnam on the mediating role of math attitude
**Original caption:** Fig. 4. : Distribution of Math performance by region, school type, and geographic location.
**Purpose:** Achievement distributions compared across region, school type and geographic location. **Visual forms:** Density/histogram/boxplot, Composite.
**Physical source pages:** 5.
[PNG](<07_Mechanisms_and_Heterogeneity/P058-Fig04-Distribution%20of%20Math%20performance%20by%20region,%20school%20type,%20and%20geographic%20location.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=472>) · Original paper and literature note are not hosted.
### P058 - Fig. 5
**Paper:** Can inquiry-based pedagogy improve math performance? Evidence from 5711 students in Vietnam on the mediating role of math attitude
**Original caption:** Fig. 5. Solution Diagram of Relationships: A visual representation illustrating the paths and standardized solutions on the interlinks of inquiry-based pedagogy, student math attitude, and student math performance (Base model).
**Purpose:** Estimated mediation pathways through student math attitude. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P058-Fig05-Solution%20Diagram%20of%20Relationships%20A%20visual%20representation%20illustrating%20the%20paths%20and%20standardized%20solutions%20on%20the%20interlinks%20of%20inquiry-based%20pedagogy,%20student....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=473>) · Original paper and literature note are not hosted.
### P059 - Fig. 3
**Paper:** Creative learning environments and mathematics self-efficacy as predictors of mathematics achievement: Insights from PISA 2022 across three countries – Korea, Canada, and Türkiye
**Original caption:** Fig. 3. Structural equation model for Korea, Canada, and Türkiye.
**Purpose:** Cross-country structural paths examining mathematics self-efficacy mediation. **Visual forms:** Path diagram.
**Physical source pages:** 12.
[PNG](<07_Mechanisms_and_Heterogeneity/P059-Fig03-Structural%20equation%20model%20for%20Korea,%20Canada,%20and%20T%C3%BCrkiye.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=474>) · Original paper and literature note are not hosted.
### P060 - Fig. 2
**Paper:** Does motivation towards science mediate the relationship between parental support and adolescent science achievement? A comparison of two countries
**Original caption:** Fig. 2. Differential Effects of Parental Support on Intrinsic Motivation.
**Purpose:** Country differences in parental-support associations with intrinsic motivation. **Visual forms:** Line/profile plot.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P060-Fig02-Differential%20Effects%20of%20Parental%20Support%20on%20Intrinsic%20Motivation.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=475>) · Original paper and literature note are not hosted.
### P060 - Fig. 3
**Paper:** Does motivation towards science mediate the relationship between parental support and adolescent science achievement? A comparison of two countries
**Original caption:** Fig. 3. Differential Effects of Parental Support on Science Achievement.
**Purpose:** Country differences in the parental-support association with achievement. **Visual forms:** Line/profile plot.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P060-Fig03-Differential%20Effects%20of%20Parental%20Support%20on%20Science%20Achievement.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=476>) · Original paper and literature note are not hosted.
### P061 - Fig. 3
**Paper:** Does the relationship between competitiveness and student achievement vary based on the culture of competitiveness? Evidence from PISA 2018
**Original caption:** Fig. 3. Significant cross-level interactions in Mathematics Achievement (Student Competitiveness and Culture of Competitiveness.
**Purpose:** Country culture moderates the student competitiveness association. **Visual forms:** Interaction plot.
**Physical source pages:** 12.
[PNG](<07_Mechanisms_and_Heterogeneity/P061-Fig03-Significant%20cross-level%20interactions%20in%20Mathematics%20Achievement%20(Student%20Competitiveness%20and%20Culture%20of%20Competitiveness.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=477>) · Original paper and literature note are not hosted.
### P062 - Fig. 3
**Paper:** PISA 2022 insights on intellectual curiosity, perspective-taking, and science achievement: Examining the mediating pathways
**Original caption:** Fig. 3. Path coefficient and R^2.
**Purpose:** Estimated mediation pathways and coefficients. **Visual forms:** Path diagram.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P062-Fig03-Path%20coefficient%20and%20R%5E2.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=478>) · Original paper and literature note are not hosted.
### P064 - Fig. 2
**Paper:** Students’ reading enjoyment as a multilevel mediator of the relationship between parental reading enjoyment, teachers’ stimulation of reading, ICT factors, and reading literacy
**Original caption:** Fig. 2. The results of multilevel structural equation modeling (standardized coefficients are presented, with p-values in parentheses).
**Purpose:** Estimated multilevel reading-enjoyment mediation. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P064-Fig02-The%20results%20of%20multilevel%20structural%20equation%20modeling%20(standardized%20coefficients%20are%20presented,%20with%20p-values%20in%20parentheses).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=479>) · Original paper and literature note are not hosted.
### P067 - Fig. 1
**Paper:** The impact of reading metacognitive strategies on mathematics learning efficiency and performance: An analysis using PISA 2018 data in China
**Original caption:** Fig. 1. Latent profile analysis of reading metacognitive strategies (Note: a = effective strategy, b = moderately effective strategy, c = ineffective strategy).
**Purpose:** Six latent metacognitive profiles compared across strategies, learning time and achievement. **Visual forms:** Bar chart.
**Physical source pages:** 4.
[PNG](<07_Mechanisms_and_Heterogeneity/P067-Fig01-Latent%20profile%20analysis%20of%20reading%20metacognitive%20strategies%20(Note%20a%20=%20effective%20strategy,%20b%20=%20moderately%20effective%20strategy,%20c%20=%20ineffective%20strategy).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=480>) · Original paper and literature note are not hosted.
### P067 - Fig. 2
**Paper:** The impact of reading metacognitive strategies on mathematics learning efficiency and performance: An analysis using PISA 2018 data in China
**Original caption:** Fig. 2. Z scores of each type of student in each variable.
**Purpose:** Item-response profiles describe heterogeneity among the six student types. **Visual forms:** Line/profile plot.
**Physical source pages:** 5.
[PNG](<07_Mechanisms_and_Heterogeneity/P067-Fig02-Z%20scores%20of%20each%20type%20of%20student%20in%20each%20variable.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=481>) · Original paper and literature note are not hosted.
### P068 - Fig. 5
**Paper:** Understanding students’ mathematical achievement through feature selection and machine learning: Evidence from PISA 2018
**Original caption:** Figure 5. PDPs, ICE plots, and ALE plots for the Random Forest model based on the top 5 features selected by Mutual Information, using PV1MATH.
**Purpose:** Nonlinear predictor-response patterns in model explanations. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 20.
[PNG](<07_Mechanisms_and_Heterogeneity/P068-Fig05-PDPs,%20ICE%20plots,%20and%20ALE%20plots%20for%20the%20Random%20Forest%20model%20based%20on%20the%20top%205%20features%20selected%20by%20Mutual%20Information,%20using%20PV1MATH.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=482>) · Original paper and literature note are not hosted.
### P076 - Fig. 1
**Paper:** The effects of teaching initiative, ICT application, and teaching collaboration on teachers’ creative thinking: Latent profile and multiple logistic regression analysis
**Original caption:** Fig. 1. Mean WLE distribution of 5 latent classes of teacher creative thinking.
**Purpose:** Latent teacher creative-thinking profiles, each with a distinct response pattern. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P076-Fig01-Mean%20WLE%20distribution%20of%205%20latent%20classes%20of%20teacher%20creative%20thinking.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=483>) · Original paper and literature note are not hosted.
### P078 - Fig. 5
**Paper:** Digital resource engagement, health challenges, and mathematical achievement in school children: An in-depth analysis of PISA data 2022
**Original caption:** Fig. 5. Effect of health problem on achievement score with mediating role of use of digital resources and confidence in mathematics learning (Model 1).
**Purpose:** Estimated mediated pathways in model 1. **Visual forms:** Path diagram.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P078-Fig05-Effect%20of%20health%20problem%20on%20achievement%20score%20with%20mediating%20role%20of%20use%20of%20digital%20resources%20and%20confidence%20in%20mathematics%20learning%20(Model%201).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=484>) · Original paper and literature note are not hosted.
### P078 - Fig. 6
**Paper:** Digital resource engagement, health challenges, and mathematical achievement in school children: An in-depth analysis of PISA data 2022
**Original caption:** Fig. 6. Effect of using digital resources and health problems on mathematics achievement with mediating role of confidence in mathematics learning (Model 2).
**Purpose:** Estimated mediated pathways in model 2. **Visual forms:** Path diagram.
**Physical source pages:** 8.
[PNG](<07_Mechanisms_and_Heterogeneity/P078-Fig06-Effect%20of%20using%20digital%20resources%20and%20health%20problems%20on%20mathematics%20achievement%20with%20mediating%20role%20of%20confidence%20in%20mathematics%20learning%20(Model%202).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=485>) · Original paper and literature note are not hosted.
### P078 - Fig. 7
**Paper:** Digital resource engagement, health challenges, and mathematical achievement in school children: An in-depth analysis of PISA data 2022
**Original caption:** Fig. 7. Effect of using digital resources on achievement score with mediating role of health problems and confidence in mathematics learning (Model 3).
**Purpose:** Estimated mediated pathways in model 3. **Visual forms:** Path diagram.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P078-Fig07-Effect%20of%20using%20digital%20resources%20on%20achievement%20score%20with%20mediating%20role%20of%20health%20problems%20and%20confidence%20in%20mathematics%20learning%20(Model%203).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=486>) · Original paper and literature note are not hosted.
### P080 - Fig. 4
**Paper:** Examining the relationship between online chatting and PISA reading literacy trends (2000-2018)
**Original caption:** Fig. 4. Empirical model – standardized path coefficients and fit indices
The statistical output yielded a TLI value greater than 1; according to convention, this is represented as 1.000.
**Purpose:** Estimated direct and mediated country-level trend pathways. **Visual forms:** Path diagram.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P080-Fig04-Empirical%20model%20-%20standardized%20path%20coefficients%20and%20fit%20indices%20The%20statistical%20output%20yielded%20a%20TLI%20value%20greater%20than%201;%20according%20to%20convention,%20this%20is....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=487>) · Original paper and literature note are not hosted.
### P081 - Fig. 1-4
**Paper:** Latent classes of students' ICT engagement behaviors: Predictive role of psychological factors based on PISA 2022 data
**Original caption:** Figs. 1-4. Latent Class Profiles of Student ICT Engagement Behaviors in Hong Kong (China), Japan, Korea, and Singapore.
**Purpose:** Four-country latent-class profile comparisons. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P081-Fig1-4-Figs.%201-4.%20Latent%20Class%20Profiles%20of%20Student%20ICT%20Engagement%20Behaviors%20in%20Hong%20Kong%20(China),%20Japan,%20Korea,%20and%20Singapore.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=488>) · Original paper and literature note are not hosted.
### P081 - Fig. 6
**Paper:** Latent classes of students' ICT engagement behaviors: Predictive role of psychological factors based on PISA 2022 data
**Original caption:** Fig. 6. Weighted Average Scores of Student ICT Engagement Behaviors Across Hong Kong (China), Japan, Korea, and Singapore.
**Purpose:** Cross-country comparison of weighted ICT engagement behaviors. **Visual forms:** Bar chart.
**Physical source pages:** 12.
[PNG](<07_Mechanisms_and_Heterogeneity/P081-Fig06-Weighted%20Average%20Scores%20of%20Student%20ICT%20Engagement%20Behaviors%20Across%20Hong%20Kong%20(China),%20Japan,%20Korea,%20and%20Singapore.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=489>) · Original paper and literature note are not hosted.
### P082 - Table 4
**Paper:** National contexts and information and communication technology in education: A systematic review of PISA studies
**Original caption:** Table 4 Summary of country-level factors’ moderation effects. Note. Bhutoria and Aljabri (2022) did not report statistical tables. Thus, moderation effects were interpreted based on figures illustrating inverted U-shaped relationships between online activity and achievement across different IDI levels. Moreover, the minimal effects and non-significant moderation effects in Park & Weng, 2020 were excluded from this summary table.
**Purpose:** Structured cross-study comparison of country moderators, predictors, outcomes and moderation types. **Visual forms:** Complex table.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P082-Table04-Summary%20of%20country-level%20factors%E2%80%99%20moderation%20effects.%20Note.%20Bhutoria%20and%20Aljabri%20(2022)%20did%20not%20report%20statistical%20tables.%20Thus,%20moderation%20effects%20were....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=490>) · Original paper and literature note are not hosted.
### P082 - Fig. 5
**Paper:** National contexts and information and communication technology in education: A systematic review of PISA studies
**Original caption:** Fig. 5. Moderation Effect Types This figure visualizes the four types of moderation effects identified in the review, based on interactions between country- and student-level ICT factors.
**Purpose:** Synthesizes moderation patterns identified in the review findings. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P082-Fig05-Moderation%20Effect%20Types%20This%20figure%20visualizes%20the%20four%20types%20of%20moderation%20effects%20identified%20in%20the%20review,%20based%20on%20interactions%20between%20country-%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=491>) · Original paper and literature note are not hosted.
### P083 - Fig. 3
**Paper:** Profiles of academic performance and perceived digital distraction in secondary classrooms
**Original caption:** Fig. 3. SHAP Summary Plot of Feature Contributions in Binary Model A (Clu_2 vs. Clu_1).
**Purpose:** Features distinguishing two student profiles. **Visual forms:** Composite, SHAP beeswarm, Bar chart.
**Physical source pages:** 11.
[PNG](<07_Mechanisms_and_Heterogeneity/P083-Fig03-SHAP%20Summary%20Plot%20of%20Feature%20Contributions%20in%20Binary%20Model%20A%20(Clu_2%20vs.%20Clu_1).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=492>) · Original paper and literature note are not hosted.
### P083 - Fig. 4
**Paper:** Profiles of academic performance and perceived digital distraction in secondary classrooms
**Original caption:** Fig. 4. SHAP Summary Plot of Feature Contributions in Binary Model B (Clu_3 vs. Clu_1).
**Purpose:** Features distinguishing two student profiles. **Visual forms:** Composite, SHAP beeswarm, Bar chart.
**Physical source pages:** 11.
[PNG](<07_Mechanisms_and_Heterogeneity/P083-Fig04-SHAP%20Summary%20Plot%20of%20Feature%20Contributions%20in%20Binary%20Model%20B%20(Clu_3%20vs.%20Clu_1).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=493>) · Original paper and literature note are not hosted.
### P083 - Fig. 5
**Paper:** Profiles of academic performance and perceived digital distraction in secondary classrooms
**Original caption:** Fig. 5. SHAP Summary Plot of Feature Contributions in Binary Model C (Clu_4 vs. Clu_1).
**Purpose:** Features distinguishing two student profiles. **Visual forms:** Composite, SHAP beeswarm, Bar chart.
**Physical source pages:** 12.
[PNG](<07_Mechanisms_and_Heterogeneity/P083-Fig05-SHAP%20Summary%20Plot%20of%20Feature%20Contributions%20in%20Binary%20Model%20C%20(Clu_4%20vs.%20Clu_1).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=494>) · Original paper and literature note are not hosted.
### P083 - Fig. 6
**Paper:** Profiles of academic performance and perceived digital distraction in secondary classrooms
**Original caption:** Fig. 6. SHAP Summary Plot of Feature Contributions in Binary Model D (Clu_3 vs. Clu_2).
**Purpose:** Features distinguishing two student profiles. **Visual forms:** Composite, SHAP beeswarm, Bar chart.
**Physical source pages:** 12.
[PNG](<07_Mechanisms_and_Heterogeneity/P083-Fig06-SHAP%20Summary%20Plot%20of%20Feature%20Contributions%20in%20Binary%20Model%20D%20(Clu_3%20vs.%20Clu_2).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=495>) · Original paper and literature note are not hosted.
### P083 - Fig. 7
**Paper:** Profiles of academic performance and perceived digital distraction in secondary classrooms
**Original caption:** Fig. 7. SHAP Summary Plot of Feature Contributions in Binary Model E (Clu_4 vs. Clu_2).
**Purpose:** Features distinguishing two student profiles. **Visual forms:** Composite, SHAP beeswarm, Bar chart.
**Physical source pages:** 13.
[PNG](<07_Mechanisms_and_Heterogeneity/P083-Fig07-SHAP%20Summary%20Plot%20of%20Feature%20Contributions%20in%20Binary%20Model%20E%20(Clu_4%20vs.%20Clu_2).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=496>) · Original paper and literature note are not hosted.
### P083 - Fig. 8
**Paper:** Profiles of academic performance and perceived digital distraction in secondary classrooms
**Original caption:** Fig. 8. illustrates the top five overrepresented countries for each profile.
**Purpose:** Country representation differs across student profiles. **Visual forms:** Composite, Bar chart.
**Physical source pages:** 13.
[PNG](<07_Mechanisms_and_Heterogeneity/P083-Fig08-illustrates%20the%20top%20five%20overrepresented%20countries%20for%20each%20profile.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=497>) · Original paper and literature note are not hosted.
### P083 - Fig. 9
**Paper:** Profiles of academic performance and perceived digital distraction in secondary classrooms
**Original caption:** Fig. 9. Gender overrepresentation across Student profiles based on standardized residuals.
**Purpose:** Gender representation differs across student profiles. **Visual forms:** Bar chart.
**Physical source pages:** 13.
[PNG](<07_Mechanisms_and_Heterogeneity/P083-Fig09-Gender%20overrepresentation%20across%20Student%20profiles%20based%20on%20standardized%20residuals.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=498>) · Original paper and literature note are not hosted.
### P083 - Fig. 10
**Paper:** Profiles of academic performance and perceived digital distraction in secondary classrooms
**Original caption:** Fig. 10. Profile characteristics by internal resources, external environments, and demographic backgrounds.
**Purpose:** Compares characteristics of four latent profiles. **Visual forms:** Line/profile plot.
**Physical source pages:** 14.
[PNG](<07_Mechanisms_and_Heterogeneity/P083-Fig10-Profile%20characteristics%20by%20internal%20resources,%20external%20environments,%20and%20demographic%20backgrounds.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=499>) · Original paper and literature note are not hosted.
### P084 - Fig. 1
**Paper:** Socioeconomic inequalities in digital practices and outcomes: Evidence from OECD countries
**Original caption:** Fig. 1. ESCS differences in digital access and engagement across discretionary and school-related digital practices in OECD countries Note: The scale reflects standardized PISA index scores (presented in standard deviation units) centered at the OECD mean (0). Positive values indicate above-average levels on the corresponding ICT access or digital practice measure, while negative values indicate below-average levels. Countries are ordered from the smallest to the largest ESCS disparities (based on low-high ESCS differences). Country labels use ISO 3166-1 alpha-3 codes (three-letter country abbreviations). Examples include AUS = Australia, AUT = Austria, BEL = Belgium, DEU=Germany, DNK = Denmark, ESP=Spain.
**Purpose:** Compares digital practices between socioeconomic groups within OECD countries. **Visual forms:** Composite, Coefficient plot.
**Physical source pages:** 7, 8, 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P084-Fig01-ESCS%20differences%20in%20digital%20access%20and%20engagement%20across%20discretionary%20and%20school-related%20digital%20practices%20in%20OECD%20countries%20Note%20The%20scale%20reflects%20standardized....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=500>) · Original paper and literature note are not hosted.
### P084 - Fig. 2
**Paper:** Socioeconomic inequalities in digital practices and outcomes: Evidence from OECD countries
**Original caption:** Fig. 2. Simple Slopes for the Interaction Between ESCS and Leisure Digital Practices in Predicting Digital Self-Efficacy Note. Lines indicate predicted values and shaded bands represent 95% confidence intervals, based on regression model controlling for home and school digital access and practices as well as individual and school background characteristics. Solid line (—) = Bottom ESCS; dashed line (– -) = Middle ESCS; dotted line (⋯) = Top ESCS. Regression results are reported in Table A5, Panel C (Online Appendix). Because marginal plots cannot be generated from multiply imputed datasets in Stata, the figure is based on one imputed dataset and is intended for illustrative purposes.
**Purpose:** Tests socioeconomic moderation of leisure digital practice and self-efficacy association. **Visual forms:** Interaction plot.
**Physical source pages:** 14.
[PNG](<07_Mechanisms_and_Heterogeneity/P084-Fig02-Simple%20Slopes%20for%20the%20Interaction%20Between%20ESCS%20and%20Leisure%20Digital%20Practices%20in%20Predicting%20Digital%20Self-Efficacy%20Note.%20Lines%20indicate%20predicted%20values%20and%20shaded....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=501>) · Original paper and literature note are not hosted.
### P085 - Fig. 6a-e
**Paper:** The impact of digital resource utilization on student learning outcomes and self-efficacy across different economic contexts: A comparative analysis of PISA, 2022
**Original caption:** Fig. 6a. Frequency based on income level of the countries. Fig. 6b. Standard deviation based of the income level of the countries. Fig. 6c. Mean score based on the income level of the countries. Fig. 6d. Item wise mean score of SA and SE related items. Fig. 6e. Item wise mean score of AS, OS, SL, and SF related items.
**Purpose:** Compares construct and item distributions by national income context. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 10, 11.
[PNG](<07_Mechanisms_and_Heterogeneity/P085-Fig6a-e-Frequency%20based%20on%20income%20level%20of%20the%20countries.%20Fig.%206b.%20Standard%20deviation%20based%20of%20the%20income%20level%20of%20the%20countries.%20Fig.%206c.%20Mean%20score%20based%20on%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=502>) · Original paper and literature note are not hosted.
### P085 - Fig. 7a-d
**Paper:** The impact of digital resource utilization on student learning outcomes and self-efficacy across different economic contexts: A comparative analysis of PISA, 2022
**Original caption:** Fig. 7a. Results based on the sample from lower middle-income countries. Fig. 7b. Results based on the sample from upper middle-income countries. Fig. 7c. Results based on the sample from higher income countries. Fig. 7d. Results based on total sample.
**Purpose:** SEM path comparisons across income groups and the complete sample. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 11.
[PNG](<07_Mechanisms_and_Heterogeneity/P085-Fig7a-d-Results%20based%20on%20the%20sample%20from%20lower%20middle-income%20countries.%20Fig.%207b.%20Results%20based%20on%20the%20sample%20from%20upper%20middle-income%20countries.%20Fig.%207c.%20Results%20based....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=503>) · Original paper and literature note are not hosted.
### P086 - Fig. 1
**Paper:** The impact of online interaction and information technology accessibility on academic engagement among international undergraduate students in Chinese universities: The mediating effect of learning interest
**Original caption:** Fig. 1. The mediating effect of learning interest in this study.
**Purpose:** Estimated mediation paths from accessibility and online interaction through interest to engagement. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P086-Fig01-The%20mediating%20effect%20of%20learning%20interest%20in%20this%20study.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=504>) · Original paper and literature note are not hosted.
### P088 - Fig. 2
**Paper:** Understanding the relationships among ICT use, self-efficacy, and achievement in PISA 2022: A multigroup analysis featuring gender and immigrant status
**Original caption:** Fig. 2. The overall model Note. *p < .05, **p < .01, ***p < .001.
**Purpose:** Overall estimated mediation pathways via two self-efficacy dimensions. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P088-Fig02-The%20overall%20model%20Note.%201%20asterisk%20p%20less%20than%20.05,%202%20asterisks%20p%20less%20than%20.01,%203%20asterisks%20p%20less%20than%20.001.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=505>) · Original paper and literature note are not hosted.
### P088 - Fig. 3
**Paper:** Understanding the relationships among ICT use, self-efficacy, and achievement in PISA 2022: A multigroup analysis featuring gender and immigrant status
**Original caption:** Fig. 3. The model based on gender grouping Note. *p < .05, **p < .01, ***p < .001.
**Purpose:** Gender-specific pathways compare relationships for girls and boys. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P088-Fig03-The%20model%20based%20on%20gender%20grouping%20Note.%201%20asterisk%20p%20less%20than%20.05,%202%20asterisks%20p%20less%20than%20.01,%203%20asterisks%20p%20less%20than%20.001.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=506>) · Original paper and literature note are not hosted.
### P088 - Fig. 4
**Paper:** Understanding the relationships among ICT use, self-efficacy, and achievement in PISA 2022: A multigroup analysis featuring gender and immigrant status
**Original caption:** Fig. 4. The model based on immigrant grouping Note. *p < .05, **p < .01, ***p < .001.
**Purpose:** Three immigrant-status groups compared using mediation paths. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P088-Fig04-The%20model%20based%20on%20immigrant%20grouping%20Note.%201%20asterisk%20p%20less%20than%20.05,%202%20asterisks%20p%20less%20than%20.01,%203%20asterisks%20p%20less%20than%20.001.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=507>) · Original paper and literature note are not hosted.
### P089 - Fig. 1
**Paper:** Gender, mindset, and teacher support predict students’ math motivation profiles: Evidence from five countries in PISA 2022
**Original caption:** Fig. 1. Latent Profiles of Math Motivation in Germany, Korea, the U.S., Finland, and Japan.
**Purpose:** Cross-country latent motivational profile comparisons. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P089-Fig01-Latent%20Profiles%20of%20Math%20Motivation%20in%20Germany,%20Korea,%20the%20U.S.,%20Finland,%20and%20Japan.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=508>) · Original paper and literature note are not hosted.
### P090 - Fig. 4
**Paper:** Meta-affective progression in science: A cross-sectional study of junior high school students
**Original caption:** Fig. 4. Radar diagram depicting each sub-indicator's meta-affective achievement of 7th and 9th grade students.
**Purpose:** Compares affective subdimension means between grade groups. **Visual forms:** Radar plot.
**Physical source pages:** 5.
[PNG](<07_Mechanisms_and_Heterogeneity/P090-Fig04-Radar%20diagram%20depicting%20each%20sub-indicator's%20meta-affective%20achievement%20of%207th%20and%209th%20grade%20students.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=509>) · Original paper and literature note are not hosted.
### P090 - Fig. 5
**Paper:** Meta-affective progression in science: A cross-sectional study of junior high school students
**Original caption:** Fig. 5. Bayesian posterior distribution curve on meta-affective sub-indicator.
**Purpose:** Six posterior difference distributions compare grade groups. **Visual forms:** Density, Composite.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P090-Fig05-Bayesian%20posterior%20distribution%20curve%20on%20meta-affective%20sub-indicator.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=510>) · Original paper and literature note are not hosted.
### P092 - Fig. 1
**Paper:** The relationship between parental emotional support, growth mindset, psychological resilience, and subjective well-being: A latent profile analysis—Based on the analysis of PISA 2018 data from four Chinese provinces
**Original caption:** Fig. 1. Different types of SWB patterns.
**Purpose:** Three latent well-being profiles identify student heterogeneity. **Visual forms:** Line/profile plot.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P092-Fig01-Different%20types%20of%20SWB%20patterns.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=511>) · Original paper and literature note are not hosted.
### P092 - Fig. 2
**Paper:** The relationship between parental emotional support, growth mindset, psychological resilience, and subjective well-being: A latent profile analysis—Based on the analysis of PISA 2018 data from four Chinese provinces
**Original caption:** Fig. 2. Path analysis of the impact of parental emotional support on different subjective well-being groups.
**Purpose:** Mediation pathways compared across three well-being profiles. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P092-Fig02-Path%20analysis%20of%20the%20impact%20of%20parental%20emotional%20support%20on%20different%20subjective%20well-being%20groups.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=512>) · Original paper and literature note are not hosted.
### P095 - Fig. 2
**Paper:** International differences in educational equity: An assessment using the Benefit of the Doubt model
**Original caption:** Fig. 2. Equity performance of the Top 10/Bottom 10 OECD equitable educational systems — unconditional DD-BoD.
**Purpose:** Equity-dimension profiles compare the best and worst educational systems. **Visual forms:** Line/profile plot.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P095-Fig02-Equity%20performance%20of%20the%20Top%2010%20Bottom%2010%20OECD%20equitable%20educational%20systems%20-%20unconditional%20DD-BoD.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=513>) · Original paper and literature note are not hosted.
### P095 - Fig. 4
**Paper:** International differences in educational equity: An assessment using the Benefit of the Doubt model
**Original caption:** Fig. 4. Non-parametric partial regression plots for the contextual variables.
**Purpose:** Context-specific relationships explaining equity heterogeneity. **Visual forms:** Scatter, Composite.
**Physical source pages:** 13.
[PNG](<07_Mechanisms_and_Heterogeneity/P095-Fig04-Non-parametric%20partial%20regression%20plots%20for%20the%20contextual%20variables.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=514>) · Original paper and literature note are not hosted.
### P096 - Fig. 3
**Paper:** Pathways to equity: A mediation analysis of gender, SES, and mathematics achievement using PISA 2022 UK data
**Original caption:** Fig. 3. The mediator model diagram on gender, ESCS, math teacher support, growth mindset, math anxiety, and math achievement. Note. gender: gender; escs: socioeconomic status; teachsup: mathematics teacher support; grosagr: growth mindset; anxmat: mathematics anxiety; pvmath: mathematics achievement.
**Purpose:** Estimated mediator paths linking gender and SES to mathematics achievement. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P096-Fig03-The%20mediator%20model%20diagram%20on%20gender,%20ESCS,%20math%20teacher%20support,%20growth%20mindset,%20math%20anxiety,%20and%20math%20achievement.%20Note.%20gender%20gender;%20escs%20socioeconomic....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=515>) · Original paper and literature note are not hosted.
### P097 - Fig. 2
**Paper:** The role of parental involvement in adolescents’ digital reading performance
**Original caption:** Fig. 2. Interaction of the PRESUPP and the EMOSUPP¹ PRECUPP means “previous parental support in reading-related activities”, and EMOSUPP means “Parental emotional support”.
**Purpose:** Interaction of early reading support and emotional support for digital reading performance. **Visual forms:** Line/profile plot.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P097-Fig02-Interaction%20of%20the%20PRESUPP%20and%20the%20EMOSUPP1%20PRECUPP%20means%20%E2%80%9Cprevious%20parental%20support%20in%20reading-related%20activities%E2%80%9D,%20and%20EMOSUPP%20means%20%E2%80%9CParental%20emotional%20support%E2%80%9D.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=516>) · Original paper and literature note are not hosted.
### P099 - Fig. 3
**Paper:** Understanding disparities: Student and school factors associated with U.S. students’ achievement in reading, mathematics, and science
**Original caption:** Fig. 3. Compositional Effects: Comparing Individual and School-Level Parental Education Associations. Note. Green bars show within-school associations (individual-level parental education holding school-mean parental education constant; orange bars show between-school associations (school-mean prenatal education holding individual parental education constant). All estimates are from Model 4 (Contextual Effects), pooled via Rubin’s rules across 10 plausible values X 10 imputations. Error bars represent 95% confidence intervals. Between-school associations are roughly 3.7–3.8x larger than within-school associations across all three domains. The wider between-school confidence intervals reflect smaller Level−2 sample size (N = 154 schools) and the use of manifest (rather than latent) school means (Lüdtke et al., 2008; Lüdtke et al., 2011).
**Purpose:** Cross-level school composition and immigration heterogeneity display; preserves the original Figure 3 caption. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 13.
[PNG](<07_Mechanisms_and_Heterogeneity/P099-Fig03-Compositional%20Effects%20Comparing%20Individual%20and%20School-Level%20Parental%20Education%20Associations.%20Note.%20Green%20bars%20show%20within-school%20associations%20(individual-level....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=517>) · Original paper and literature note are not hosted.
### P099 - Fig. 4
**Paper:** Understanding disparities: Student and school factors associated with U.S. students’ achievement in reading, mathematics, and science
**Original caption:** Fig. 4. Predicted Achievement by School Socioeconomic Composition and Immigration Status (Cross-Level Interaction Model). Note. Predicted values derived from Model 5 (Cross-Level Interactions), with other predictors held at reference values: Female for gender, English for home language, the smallest community category for community type, and sample means for parental education, home ICT resources, and school ICT availability. Lines are parallel across immigration status, reflecting the non-significant interactions between immigration status and school socioeconomic composition (all p > .20 across the domains).
**Purpose:** School-composition and immigration interaction/within-between decomposition display; caption is preserved exactly despite source mismatch. **Visual forms:** Bar chart.
**Physical source pages:** 13.
[PNG](<07_Mechanisms_and_Heterogeneity/P099-Fig04-Predicted%20Achievement%20by%20School%20Socioeconomic%20Composition%20and%20Immigration%20Status%20(Cross-Level%20Interaction%20Model).%20Note.%20Predicted%20values%20derived%20from%20Model%205....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=518>) · Original paper and literature note are not hosted.
### P099 - Fig. 5
**Paper:** Understanding disparities: Student and school factors associated with U.S. students’ achievement in reading, mathematics, and science
**Original caption:** Fig. 5. School-Specific Variation in Gender and Language Effects (Random Slopes). Note. Violin and boxplot distributions represent school-specific effects (the sum of each school’s estimated random slope and the corresponding pooled fixed effect) across 154 schools, drawn from a representative fit of Model 6 (Random Slopes). Dashed vertical lines indicate the pooled fixed-effect estimate for each domain. The wide spread of school-specific effects documents substantial heterogeneity in how gender and home language relate to achievement across U.S. schools; across-fit variation is small relative to the displayed school-level heterogeneity.
**Purpose:** School-specific heterogeneity in gender/language random slopes. **Visual forms:** Density/histogram/boxplot, Composite.
**Physical source pages:** 14.
[PNG](<07_Mechanisms_and_Heterogeneity/P099-Fig05-School-Specific%20Variation%20in%20Gender%20and%20Language%20Effects%20(Random%20Slopes).%20Note.%20Violin%20and%20boxplot%20distributions%20represent%20school-specific%20effects%20(the%20sum%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=519>) · Original paper and literature note are not hosted.
### P100 - Fig. 1
**Paper:** When do educational expectations motivate effort? Expectation-opportunity alignment and study time across 27 countries
**Original caption:** Fig. 1. The association between expectations and study time, by country. Note: Countries are ranked in descending order of the unconditional coefficient of HE expectation. Hollow bars and circles represent statistically insignificant parameters at the 5% level. Controls include gender, PISA math score and its squared term, school track and level, school instruction time, SES quartile, immigration status, and school fixed effect; the outcome variable (study time) is standardized within countries.
**Purpose:** Country heterogeneity of the expectations–study-time association. **Visual forms:** Bar chart, Scatter.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P100-Fig01-The%20association%20between%20expectations%20and%20study%20time,%20by%20country.%20Note%20Countries%20are%20ranked%20in%20descending%20order%20of%20the%20unconditional%20coefficient%20of%20HE%20expectation....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=520>) · Original paper and literature note are not hosted.
### P100 - Fig. 2
**Paper:** When do educational expectations motivate effort? Expectation-opportunity alignment and study time across 27 countries
**Original caption:** Fig. 2. The marginal effect of expectation on study time, by EOA index Note: The graph is based on Model 3 from Table 2. Confidence intervals are at the 95% level.
**Purpose:** Nonlinear national EOA moderation of the individual expectation–effort association. **Visual forms:** Line/profile plot.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P100-Fig02-The%20marginal%20effect%20of%20expectation%20on%20study%20time,%20by%20EOA%20index%20Note%20The%20graph%20is%20based%20on%20Model%203%20from%20Table%202.%20Confidence%20intervals%20are%20at%20the%2095%25%20level.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=521>) · Original paper and literature note are not hosted.
### P102 - Fig. 2
**Paper:** Fostering school and classroom discipline in Türkiye: Leadership, shared decision-making, parental involvement and teacher self-efficacy
**Original caption:** Fig. 2. Structural model showing direct paths among variables. Source: Authors own works.
**Purpose:** Estimated direct and mediated relations among school and classroom factors. **Visual forms:** Path diagram.
**Physical source pages:** 8.
[PNG](<07_Mechanisms_and_Heterogeneity/P102-Fig02-Structural%20model%20showing%20direct%20paths%20among%20variables.%20Source%20Authors%20own%20works.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=522>) · Original paper and literature note are not hosted.
### P103 - Fig. 2
**Paper:** Impact of school leadership on teachers’ instructional practices: examining mediating roles of collegiality and teacher autonomy using Vietnam TALIS 2018
**Original caption:** Fig. 2. Study model diagram with parameter estimates.
**Purpose:** Estimated multilevel direct and mediated leadership associations. **Visual forms:** Path diagram.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P103-Fig02-Study%20model%20diagram%20with%20parameter%20estimates.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=523>) · Original paper and literature note are not hosted.
### P105 - Fig. 2
**Paper:** Overworked and under-creative: Teacher workload, time allocation, and creative pedagogy in Chinese secondary schools - evidence from PISA 2022
**Original caption:** Fig. 2. Marginal effects.
**Purpose:** Six panels compare nonlinear time-allocation relationships and moderation by digital resources. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P105-Fig02-Marginal%20effects.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=524>) · Original paper and literature note are not hosted.
### P108 - Fig. 2
**Paper:** Understanding the relationship between teacher collaboration and instructional clarity via teacher self-efficacy: A moderated mediation model of organizational commitment
**Original caption:** Fig. 2. MSEM Diagram with Standardized Results (RQ1).
**Purpose:** Estimated multilevel collaboration–efficacy–instructional-clarity mediation. **Visual forms:** Path diagram.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P108-Fig02-MSEM%20Diagram%20with%20Standardized%20Results%20(RQ1).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=525>) · Original paper and literature note are not hosted.
### P108 - Fig. 3
**Paper:** Understanding the relationship between teacher collaboration and instructional clarity via teacher self-efficacy: A moderated mediation model of organizational commitment
**Original caption:** Fig. 3. Multilevel LMS Diagram with Unstandardized Results (RQ2).
**Purpose:** Estimated organizational-commitment moderation of mediation. **Visual forms:** Path diagram.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P108-Fig03-Multilevel%20LMS%20Diagram%20with%20Unstandardized%20Results%20(RQ2).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=526>) · Original paper and literature note are not hosted.
### P108 - Fig. 4
**Paper:** Understanding the relationship between teacher collaboration and instructional clarity via teacher self-efficacy: A moderated mediation model of organizational commitment
**Original caption:** Fig. 4. Organizational Commitment as a Moderator of the Effects between Teacher Collaboration and the Clarity of Instruction through Teacher Self-efficacy in Instruction. Functions are plotted for three levels of Organizational Commitment (OC): Low, Medium, and High.
**Purpose:** Conditional slopes compare the mediated association at three commitment levels. **Visual forms:** Line/profile plot.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P108-Fig04-Organizational%20Commitment%20as%20a%20Moderator%20of%20the%20Effects%20between%20Teacher%20Collaboration%20and%20the%20Clarity%20of%20Instruction%20through%20Teacher%20Self-efficacy%20in%20Instruction....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=527>) · Original paper and literature note are not hosted.
### P109 - Fig. 1
**Paper:** Associations between time spent communicating with parents, teacher self-efficacy, and stress: The role of professional development
**Original caption:** Fig. 1. Associations between time spent communicating with parents and teacher self-efficacy and parent-related stress across countries.
**Purpose:** Compares country-specific communication associations for self-efficacy and stress. **Visual forms:** Composite, Forest/coefficient plot.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P109-Fig01-Associations%20between%20time%20spent%20communicating%20with%20parents%20and%20teacher%20self-efficacy%20and%20parent-related%20stress%20across%20countries.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=528>) · Original paper and literature note are not hosted.
### P110 - Fig. 3
**Paper:** Attrition risk profiles of novice teachers based on their job demands, individual and school resources, and well-being
**Original caption:** Fig. 3. Attrition risk profiles with new labels based on a summary of the job demands, school resources, individual resources, and well-being variables (z-scores on the y-axis).
**Purpose:** Attrition-risk profiles contrasted across demands, resources and well-being. **Visual forms:** Bar chart.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P110-Fig03-Attrition%20risk%20profiles%20with%20new%20labels%20based%20on%20a%20summary%20of%20the%20job%20demands,%20school%20resources,%20individual%20resources,%20and%20well-being%20variables%20(z-scores%20on%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=529>) · Original paper and literature note are not hosted.
### P111 - Fig. 2
**Paper:** Distributed leadership and teacher job satisfaction over time: A large-scale trend analysis of TALIS 2013 and 2018
**Original caption:** Fig. 2. Mean job satisfaction with school environment (TJSENV) in 2013 and 2018 by country. Note: Each panel shows the mean values of teacher job satisfaction with their school environment for 2013 and 2018, based on TALIS data. Only countries with available data for both years are included. Δ values indicate the mean difference between 2018 and 2013, with statistical significance denoted as follows: p < .05 (*), p < .01 (**), p < .001 (***), NA indicates non-significant.
**Purpose:** Country-specific changes in environmental job satisfaction. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P111-Fig02-Mean%20job%20satisfaction%20with%20school%20environment%20(TJSENV)%20in%202013%20and%202018%20by%20country.%20Note%20Each%20panel%20shows%20the%20mean%20values%20of%20teacher%20job%20satisfaction%20with%20their....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=530>) · Original paper and literature note are not hosted.
### P111 - Fig. 3
**Paper:** Distributed leadership and teacher job satisfaction over time: A large-scale trend analysis of TALIS 2013 and 2018
**Original caption:** Fig. 3. Mean job satisfaction with profession (TJSPRO) in 2013 and 2018 by country. Note: Each panel shows the mean values of teacher job satisfaction with their profession for 2013 and 2018, based on TALIS data. Only countries with available data for both years are included. Δ values indicate the mean difference between 2018 and 2013, with statistical significance denoted as follows: p < .05 (*), p < .01 (**), p < .001 (***), NA indicates non-significant.
**Purpose:** Country-specific changes in professional job satisfaction. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P111-Fig03-Mean%20job%20satisfaction%20with%20profession%20(TJSPRO)%20in%202013%20and%202018%20by%20country.%20Note%20Each%20panel%20shows%20the%20mean%20values%20of%20teacher%20job%20satisfaction%20with%20their....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=531>) · Original paper and literature note are not hosted.
### P111 - Fig. 4
**Paper:** Distributed leadership and teacher job satisfaction over time: A large-scale trend analysis of TALIS 2013 and 2018
**Original caption:** Fig. 4. Path a (DISTLEAD → JSENV) over time by country. Note: This figure presents standardized estimates for the path from perceived distributed leadership (DISTLEAD) to job satisfaction with the school environment (JSENV) across 2013 and 2018, disaggregated by country. Only countries with data available for both years are included. Each panel illustrates the change in path strength over time, based on structural equation modeling results.
**Purpose:** Country heterogeneity in changes in leadership–environmental satisfaction paths. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 8.
[PNG](<07_Mechanisms_and_Heterogeneity/P111-Fig04-Path%20a%20(DISTLEAD%20%E2%86%92%20JSENV)%20over%20time%20by%20country.%20Note%20This%20figure%20presents%20standardized%20estimates%20for%20the%20path%20from%20perceived%20distributed%20leadership%20(DISTLEAD)%20to....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=532>) · Original paper and literature note are not hosted.
### P111 - Fig. 5
**Paper:** Distributed leadership and teacher job satisfaction over time: A large-scale trend analysis of TALIS 2013 and 2018
**Original caption:** Fig. 5. Path b (DISTLEAD → JSPRO) over time by country. Note: This figure displays standardized estimates for the path from perceived distributed leadership (DISTLEAD) to job satisfaction with the profession (JSPRO) across 2013 and 2018, separately by country. Only countries with available data for both years are shown. Each panel illustrates how the strength of the relationship evolved over time, based on structural equation modeling results.
**Purpose:** Country heterogeneity in changes in leadership–professional satisfaction paths. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P111-Fig05-Path%20b%20(DISTLEAD%20%E2%86%92%20JSPRO)%20over%20time%20by%20country.%20Note%20This%20figure%20displays%20standardized%20estimates%20for%20the%20path%20from%20perceived%20distributed%20leadership%20(DISTLEAD)%20to....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=533>) · Original paper and literature note are not hosted.
### P113 - Fig. 2
**Paper:** How perceived teaching autonomy predicts job satisfaction in China: The mediating role of self-efficacy and the moderating role of school climate
**Original caption:** Fig. 2. Moderated mediation effect.
**Purpose:** Estimated moderated-mediation paths. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P113-Fig02-Moderated%20mediation%20effect.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=534>) · Original paper and literature note are not hosted.
### P113 - Fig. 3
**Paper:** How perceived teaching autonomy predicts job satisfaction in China: The mediating role of self-efficacy and the moderating role of school climate
**Original caption:** Fig. 3. The moderating effect of school climate on the relationship between teaching autonomy and job satisfaction.
**Purpose:** Simple slopes for school-climate moderation of autonomy–satisfaction association. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P113-Fig03-The%20moderating%20effect%20of%20school%20climate%20on%20the%20relationship%20between%20teaching%20autonomy%20and%20job%20satisfaction.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=535>) · Original paper and literature note are not hosted.
### P113 - Fig. 4
**Paper:** How perceived teaching autonomy predicts job satisfaction in China: The mediating role of self-efficacy and the moderating role of school climate
**Original caption:** Fig. 4. The moderating effect of school climate on teaching autonomy in predicting self-efficacy.
**Purpose:** Simple slopes for school-climate moderation of autonomy–self-efficacy association. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P113-Fig04-The%20moderating%20effect%20of%20school%20climate%20on%20teaching%20autonomy%20in%20predicting%20self-efficacy.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=536>) · Original paper and literature note are not hosted.
### P113 - Fig. 5
**Paper:** How perceived teaching autonomy predicts job satisfaction in China: The mediating role of self-efficacy and the moderating role of school climate
**Original caption:** Fig. 5. The moderating effect of school climate on the prediction of job satisfaction by self-efficacy.
**Purpose:** Simple slopes for school-climate moderation of self-efficacy–satisfaction association. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P113-Fig05-The%20moderating%20effect%20of%20school%20climate%20on%20the%20prediction%20of%20job%20satisfaction%20by%20self-efficacy.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=537>) · Original paper and literature note are not hosted.
### P114 - Fig. 2
**Paper:** Investigating the relationship between instructional leadership and teacher well-being: A moderated mediation analysis of PISA 2022
**Original caption:** Fig. 2. The moderating effect of perceived autonomy support on teachers’ workload stress and teacher well-being.
**Purpose:** Fifteen jurisdictions compare workload-stress slopes at low and high autonomy. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P114-Fig02-The%20moderating%20effect%20of%20perceived%20autonomy%20support%20on%20teachers%E2%80%99%20workload%20stress%20and%20teacher%20well-being.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=538>) · Original paper and literature note are not hosted.
### P116 - Fig. 4
**Paper:** School leadership and teacher stress: Unpacking the mediating role of job satisfaction and the employment status as a moderator
**Original caption:** Fig. 4. Path Analysis.
**Purpose:** Estimated mediation/moderation path results. **Visual forms:** Path diagram.
**Physical source pages:** 8.
[PNG](<07_Mechanisms_and_Heterogeneity/P116-Fig04-Path%20Analysis.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=539>) · Original paper and literature note are not hosted.
### P117 - Fig. 2
**Paper:** The effect of teacher collaboration on teachers' career well-being in China: A moderated mediation model of teacher self-efficacy and distributed leadership
**Original caption:** Fig. 2. The effects of teacher collaboration on teachers’ career well-being.
**Purpose:** Estimated indirect pathway through self-efficacy. **Visual forms:** Path diagram.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P117-Fig02-The%20effects%20of%20teacher%20collaboration%20on%20teachers%E2%80%99%20career%20well-being.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=540>) · Original paper and literature note are not hosted.
### P117 - Fig. 3
**Paper:** The effect of teacher collaboration on teachers' career well-being in China: A moderated mediation model of teacher self-efficacy and distributed leadership
**Original caption:** Fig. 3. Distributed leadership moderates the relationship between teacher collaboration and teachers' career well-being. TC: teacher collaboration, DL: distributed leadership.
**Purpose:** Distributed leadership moderates the collaboration–well-being association. **Visual forms:** Interaction plot.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P117-Fig03-Distributed%20leadership%20moderates%20the%20relationship%20between%20teacher%20collaboration%20and%20teachers'%20career%20well-being.%20TC%20teacher%20collaboration,%20DL%20distributed%20leadership.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=541>) · Original paper and literature note are not hosted.
### P117 - Fig. 4
**Paper:** The effect of teacher collaboration on teachers' career well-being in China: A moderated mediation model of teacher self-efficacy and distributed leadership
**Original caption:** Fig. 4. Distributed leadership moderates the relationship between teacher collaboration and teachers' self-efficacy. TC: teacher collaboration, DL: distributed leadership.
**Purpose:** Distributed leadership moderates the collaboration–self-efficacy association. **Visual forms:** Interaction plot.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P117-Fig04-Distributed%20leadership%20moderates%20the%20relationship%20between%20teacher%20collaboration%20and%20teachers'%20self-efficacy.%20TC%20teacher%20collaboration,%20DL%20distributed%20leadership.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=542>) · Original paper and literature note are not hosted.
### P118 - Fig. 3
**Paper:** Understanding the relationship between teacher collaboration beliefs and job satisfaction via collaborative attitudes: A moderated mediation model of classroom self-efficacy from PISA 2022
**Original caption:** Fig. 3. The direct and indirect effects results in the equation model.
**Purpose:** Estimated direct and collaborative-attitude-mediated associations. **Visual forms:** Path diagram.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P118-Fig03-The%20direct%20and%20indirect%20effects%20results%20in%20the%20equation%20model.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=543>) · Original paper and literature note are not hosted.
### P118 - Fig. 4
**Paper:** Understanding the relationship between teacher collaboration beliefs and job satisfaction via collaborative attitudes: A moderated mediation model of classroom self-efficacy from PISA 2022
**Original caption:** Fig. 4. The mediation of collaborative attitudes moderated by classroom self-efficacy.
**Purpose:** Classroom self-efficacy moderation of mediation. **Visual forms:** Path diagram.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P118-Fig04-The%20mediation%20of%20collaborative%20attitudes%20moderated%20by%20classroom%20self-efficacy.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=544>) · Original paper and literature note are not hosted.
### P118 - Fig. 5
**Paper:** Understanding the relationship between teacher collaboration beliefs and job satisfaction via collaborative attitudes: A moderated mediation model of classroom self-efficacy from PISA 2022
**Original caption:** Fig. 5. The mediation of collaborative attitudes moderated by gender (Female teachers).
**Purpose:** Conditional mediation for female teachers. **Visual forms:** Path diagram.
**Physical source pages:** 11.
[PNG](<07_Mechanisms_and_Heterogeneity/P118-Fig05-The%20mediation%20of%20collaborative%20attitudes%20moderated%20by%20gender%20(Female%20teachers).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=545>) · Original paper and literature note are not hosted.
### P118 - Fig. 6
**Paper:** Understanding the relationship between teacher collaboration beliefs and job satisfaction via collaborative attitudes: A moderated mediation model of classroom self-efficacy from PISA 2022
**Original caption:** Fig. 6. The mediation of collaborative attitudes moderated by gender (Male teachers).
**Purpose:** Conditional mediation for male teachers. **Visual forms:** Path diagram.
**Physical source pages:** 11.
[PNG](<07_Mechanisms_and_Heterogeneity/P118-Fig06-The%20mediation%20of%20collaborative%20attitudes%20moderated%20by%20gender%20(Male%20teachers).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=546>) · Original paper and literature note are not hosted.
### P121 - Fig. 1_a
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 1_a. X-bar and R control chart of teachers’ competencies based on before and after COVID-19.
**Purpose:** Compares competency distributions and variability across pre/post-pandemic groups. **Visual forms:** Composite, Control chart.
**Physical source pages:** 5.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig1_a-X-bar%20and%20R%20control%20chart%20of%20teachers%E2%80%99%20competencies%20based%20on%20before%20and%20after%20COVID-19.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=547>) · Original paper and literature note are not hosted.
### P121 - Fig. 1_b
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 1_b. X-bar and R control chart of teachers’ competencies based on country income level.
**Purpose:** Compares competency distributions and variability across income contexts. **Visual forms:** Composite, Control chart.
**Physical source pages:** 5.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig1_b-X-bar%20and%20R%20control%20chart%20of%20teachers%E2%80%99%20competencies%20based%20on%20country%20income%20level.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=548>) · Original paper and literature note are not hosted.
### P121 - Fig. 2
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 2. Mean of empowerment, adaptability, management, professionalism, inclusion, and collaboration based on before and after COVID-19 pandemic.
**Purpose:** Compares teacher competencies before/after the pandemic. **Visual forms:** Bar chart.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig02-Mean%20of%20empowerment,%20adaptability,%20management,%20professionalism,%20inclusion,%20and%20collaboration%20based%20on%20before%20and%20after%20COVID-19%20pandemic.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=549>) · Original paper and literature note are not hosted.
### P121 - Fig. 3
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 3. Mean of empowerment, adaptability, management, professionalism, inclusion, and collaboration based economic level of the country.
**Purpose:** Compares teacher competencies across economic contexts. **Visual forms:** Bar chart.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig03-Mean%20of%20empowerment,%20adaptability,%20management,%20professionalism,%20inclusion,%20and%20collaboration%20based%20economic%20level%20of%20the%20country.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=550>) · Original paper and literature note are not hosted.
### P121 - Fig. 4_a
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 4_a. Centrality Plot based on income level of the country.
**Purpose:** Compares network centrality across income contexts. **Visual forms:** Composite, Line/profile plot.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig4_a-Centrality%20Plot%20based%20on%20income%20level%20of%20the%20country.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=551>) · Original paper and literature note are not hosted.
### P121 - Fig. 4_b
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 4_b. Centrality Plot based on results of before and after COVID-19.
**Purpose:** Compares network centrality before/after the pandemic. **Visual forms:** Composite, Line/profile plot.
**Physical source pages:** 8.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig4_b-Centrality%20Plot%20based%20on%20results%20of%20before%20and%20after%20COVID-19.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=552>) · Original paper and literature note are not hosted.
### P121 - Fig. 4_c
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 4_c. Clustering Plot based on income level of the country.
**Purpose:** Compares network clustering across income contexts. **Visual forms:** Composite, Line/profile plot.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig4_c-Clustering%20Plot%20based%20on%20income%20level%20of%20the%20country.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=553>) · Original paper and literature note are not hosted.
### P121 - Fig. 4_d
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 4_d. Clustering Plot based on results of before and after COVID-19.
**Purpose:** Group-specific competency relationships. **Visual forms:** Composite, Line/profile plot.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig4_d-Clustering%20Plot%20based%20on%20results%20of%20before%20and%20after%20COVID-19.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=554>) · Original paper and literature note are not hosted.
### P121 - Fig. 4_e
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 4_e. Flexplot based on country income level in relationship between competencies and PR.
**Purpose:** Compares competency–professionalism relationships across income contexts. **Visual forms:** Composite, Scatter.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig4_e-Flexplot%20based%20on%20country%20income%20level%20in%20relationship%20between%20competencies%20and%20PR.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=555>) · Original paper and literature note are not hosted.
### P121 - Fig. 4_f
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 4_f. Flexplot based on country income level in relationship between competencies and IN.
**Purpose:** Group-specific competency relationships. **Visual forms:** Composite, Scatter.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig4_f-Flexplot%20based%20on%20country%20income%20level%20in%20relationship%20between%20competencies%20and%20IN.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=556>) · Original paper and literature note are not hosted.
### P121 - Fig. 4_g
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 4_g. Flexplot based on country income level in relationship between competencies and CI.
**Purpose:** Compares competency–collaboration relationships across income contexts. **Visual forms:** Composite, Scatter.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig4_g-Flexplot%20based%20on%20country%20income%20level%20in%20relationship%20between%20competencies%20and%20CI.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=557>) · Original paper and literature note are not hosted.
### P121 - Fig. 4_h
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 4_h. Flexplot based on before and after COVID-19 in relationship between competencies and PR.
**Purpose:** Group-specific competency relationships. **Visual forms:** Composite, Scatter.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig4_h-Flexplot%20based%20on%20before%20and%20after%20COVID-19%20in%20relationship%20between%20competencies%20and%20PR.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=558>) · Original paper and literature note are not hosted.
### P121 - Fig. 4_i
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 4_i. Flexplot based on before and after COVID-19 in relationship between competencies and IN.
**Purpose:** Compares competency–inclusion relationships before/after the pandemic. **Visual forms:** Composite, Scatter.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig4_i-Flexplot%20based%20on%20before%20and%20after%20COVID-19%20in%20relationship%20between%20competencies%20and%20IN.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=559>) · Original paper and literature note are not hosted.
### P121 - Fig. 4_j
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 4_j. Flexplot based on before and after COVID-19 in relationship between competencies and CI.
**Purpose:** Compares competency–collaboration relationships before/after the pandemic. **Visual forms:** Composite, Scatter.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig4_j-Flexplot%20based%20on%20before%20and%20after%20COVID-19%20in%20relationship%20between%20competencies%20and%20CI.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=560>) · Original paper and literature note are not hosted.
### P121 - Fig. 5
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 5. Effect of needs of professional development and collaboration on competencies.
**Purpose:** Estimated mediation paths involving development needs and collaboration. **Visual forms:** Path diagram.
**Physical source pages:** 11.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig05-Effect%20of%20needs%20of%20professional%20development%20and%20collaboration%20on%20competencies.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=561>) · Original paper and literature note are not hosted.
### P121 - Fig. 6
**Paper:** Navigating the impact of the pandemic on teacher competencies: A comparative analysis across economic contexts using pre- and post-COVID-19 PISA data
**Original caption:** Fig. 6. Results based on pre-and post-COVID-19 and country economy.
**Purpose:** Estimated paths compare pandemic and economic contexts. **Visual forms:** Path diagram.
**Physical source pages:** 11.
[PNG](<07_Mechanisms_and_Heterogeneity/P121-Fig06-Results%20based%20on%20pre-and%20post-COVID-19%20and%20country%20economy.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=562>) · Original paper and literature note are not hosted.
### P124 - Fig. 2
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** Fig. 2. PLC by country.
**Purpose:** Country heterogeneity in professional-learning-community associations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P124-Fig02-PLC%20by%20country.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=563>) · Original paper and literature note are not hosted.
### P124 - Fig. 3
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** Fig. 3. Shared vision and responsibilities by country.
**Purpose:** Country heterogeneity in shared-vision associations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 8.
[PNG](<07_Mechanisms_and_Heterogeneity/P124-Fig03-Shared%20vision%20and%20responsibilities%20by%20country.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=564>) · Original paper and literature note are not hosted.
### P124 - Fig. 4
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** Fig. 4. Supportive conditions by country.
**Purpose:** Country heterogeneity in supportive-conditions associations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P124-Fig04-Supportive%20conditions%20by%20country.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=565>) · Original paper and literature note are not hosted.
### P124 - Fig. 5
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** Fig. 5. Collaborative practice with focus on student learning by country.
**Purpose:** Country heterogeneity in collaborative-practice associations. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 11.
[PNG](<07_Mechanisms_and_Heterogeneity/P124-Fig05-Collaborative%20practice%20with%20focus%20on%20student%20learning%20by%20country.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=566>) · Original paper and literature note are not hosted.
### P124 - Fig. 6
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** Fig. 6. Subgroup analysis, pooled data. Note: PLC: Professional learning community., SVR: Shared vision and responsibilities. SC: Supportive conditions. CPL: Collaborative practice with focus on student learning.
**Purpose:** Pooled subgroup differences by sex and teaching experience. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 12.
[PNG](<07_Mechanisms_and_Heterogeneity/P124-Fig06-Subgroup%20analysis,%20pooled%20data.%20Note%20PLC%20Professional%20learning%20community.,%20SVR%20Shared%20vision%20and%20responsibilities.%20SC%20Supportive%20conditions.%20CPL%20Collaborative....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=567>) · Original paper and literature note are not hosted.
### P126 - Fig. 1
**Paper:** Research on the influence of teachers’ ICT Capacity on classroom teaching quality: An intermediary model based on self-efficacy
**Original caption:** Fig. 1. Comparison of Shanghai and OECD Teachers’ ICT capacity.
**Purpose:** Shanghai versus OECD comparison of teachers ICT capacity. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P126-Fig01-Comparison%20of%20Shanghai%20and%20OECD%20Teachers%E2%80%99%20ICT%20capacity.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=568>) · Original paper and literature note are not hosted.
### P131 - Fig. 4
**Paper:** Educational equity measurement: A comparative analysis of equity indicators
**Original caption:** Fig. 4. Standardized RMSE of ESCS R² Equity Estimates Under Four Measurement Error Conditions, by Mean Reading Achievement Level. Note. Panel (a) compares small vs. Large SEM conditions (RMSE universally increases with higher measurement error). Panel (b) contrasts easy vs. difficult test conditions, revealing an inverse pattern: countries with lower mean reading literacy (left) show higher RMSE under difficult tests, whereas higher-achievement countries (right) exhibit increased RMSE with easier tests.
**Purpose:** Measurement-error consequences differ across achievement levels. **Visual forms:** Scatter, Composite.
**Physical source pages:** 12.
[PNG](<07_Mechanisms_and_Heterogeneity/P131-Fig04-Standardized%20RMSE%20of%20ESCS%20R2%20Equity%20Estimates%20Under%20Four%20Measurement%20Error%20Conditions,%20by%20Mean%20Reading%20Achievement%20Level.%20Note.%20Panel%20(a)%20compares%20small%20vs.%20Large....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=569>) · Original paper and literature note are not hosted.
### P132 - Fig. 2
**Paper:** Evaluating test-taking motivation based on process data and its influence on academic performance
**Original caption:** Fig. 2. The interactive effects of three types of effort and culture on science academic performance.
**Purpose:** Interactions of effort type and cultural grouping with performance. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 8.
[PNG](<07_Mechanisms_and_Heterogeneity/P132-Fig02-The%20interactive%20effects%20of%20three%20types%20of%20effort%20and%20culture%20on%20science%20academic%20performance.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=570>) · Original paper and literature note are not hosted.
### P134 - Fig. 8
**Paper:** Separating prior knowledge from acquired knowledge: An individual differences analysis of PISA - learning in the digital world
**Original caption:** Fig. 8. Path Model showing the Effect of Mastery Orientation, mediated by Use of Worked Examples.
**Purpose:** Estimated mastery-orientation mediation through worked-example use. **Visual forms:** Path diagram.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P134-Fig08-Path%20Model%20showing%20the%20Effect%20of%20Mastery%20Orientation,%20mediated%20by%20Use%20of%20Worked%20Examples.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=571>) · Original paper and literature note are not hosted.
### P136 - Fig. 5
**Paper:** The essence of creative thinking in PISA 2022: An item-level machine learning approach to construct validity at the facet level
**Original caption:** Fig. 5. Estimated relationship between abilities and scoring categories conditional on the domain. Note. GDI = generate diverse ideas, GCI = generate creative ideas, EII = elaborate and improve ideas, ScPS = science, SoPS = social, VE = verbal expression, WE = written expression.
**Purpose:** Ability/scoring relations differ by creativity task domain. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 13.
[PNG](<07_Mechanisms_and_Heterogeneity/P136-Fig05-Estimated%20relationship%20between%20abilities%20and%20scoring%20categories%20conditional%20on%20the%20domain.%20Note.%20GDI%20=%20generate%20diverse%20ideas,%20GCI%20=%20generate%20creative%20ideas,%20EII....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=572>) · Original paper and literature note are not hosted.
### P136 - Fig. 6
**Paper:** The essence of creative thinking in PISA 2022: An item-level machine learning approach to construct validity at the facet level
**Original caption:** Fig. 6. Estimated relationship between abilities and expected item scores conditional on the domain. Note. GDI = generate diverse ideas, GCI = generate creative ideas, EII = elaborate and improve ideas, ScPS = science, SoPS = social, VE = verbal expression, WE = written expression.
**Purpose:** Domain heterogeneity in expected-score relationships. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 14.
[PNG](<07_Mechanisms_and_Heterogeneity/P136-Fig06-Estimated%20relationship%20between%20abilities%20and%20expected%20item%20scores%20conditional%20on%20the%20domain.%20Note.%20GDI%20=%20generate%20diverse%20ideas,%20GCI%20=%20generate%20creative%20ideas,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=573>) · Original paper and literature note are not hosted.
### P137 - Fig. 4
**Paper:** The role of item format in the PISA 2018 mathematics literacy assessment: A cross-country study
**Original caption:** Fig. 4. Interaction Term b12 Between Item Format and Gender Obtained from Multivariate Meta-Analysis. Note. The interaction term regression weights b12 from the interaction models (M 3) are provided on the logit scale. Positive and negative b12 values indicate, respectively, smaller and larger differences in the probability of a correct response between multiple-choice (MC) vs constructed response (CR) items for male students, as compared to female students with the same mathematics ability. The results of the multivariate random effects meta-analysis are displayed at the bottom, with the black diamond representing the 95 % confidence interval of the pooled average b12, while the surrounding bars illustrate the prediction interval, reflecting the dispersion around this pooled average. Countries that administered the alternative M6B cluster are denoted with an asterisk (*).
**Purpose:** Country-specific format-by-gender interaction estimates. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 8.
[PNG](<07_Mechanisms_and_Heterogeneity/P137-Fig04-Interaction%20Term%20b12%20Between%20Item%20Format%20and%20Gender%20Obtained%20from%20Multivariate%20Meta-Analysis.%20Note.%20The%20interaction%20term%20regression%20weights%20b12%20from%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=574>) · Original paper and literature note are not hosted.
### P137 - Fig. C1
**Paper:** The role of item format in the PISA 2018 mathematics literacy assessment: A cross-country study
**Original caption:** Fig. C1. Differences in Probability of a Correct Response Between Multiple-Choice (MC) and Constructed Response (CR) for Female and Male Students. Note. ΔP represents the difference in the probability of a correct response between multiple-choice (MC) and constructed response (CR) formats’ items across genders (female or male). Probabilities for each format were first calculated separately for each gender, followed by a computation of the differences between items of two formats within each gender. The estimates are displayed in black if the interaction term b12 was significant (see Fig. 4), grey otherwise. Countries are listed in ascending order based on the gender-related difference in item format. An asterisk (*) marks countries that administered the alternative M6B cluster.
**Purpose:** Gender heterogeneity in item-format response probabilities. **Visual forms:** Scatter, Composite.
**Physical source pages:** 12.
[PNG](<07_Mechanisms_and_Heterogeneity/P137-FigC1-Differences%20in%20Probability%20of%20a%20Correct%20Response%20Between%20Multiple-Choice%20(MC)%20and%20Constructed%20Response%20(CR)%20for%20Female%20and%20Male%20Students.%20Note.%20%CE%94P%20represents%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=575>) · Original paper and literature note are not hosted.
### P141 - Fig. 4
**Paper:** Unraveling the nexus: Culture, cognitive competence, and economic performance across 86 nations (2000–2018)
**Original caption:** Fig. 4. Effects of Three Teachings on economic growth directly and mediated via PISA scores. Notes. Standardized path coefficients (β) with correlations (r) in parentheses; unexplained variance on the right; control variables are PISA (means, span, tilts), Geo. Latitude, Geo. Longitude and Log. GDP/c PPP (calculated total effects on the paths); all results based on Jackknife-technique.
**Purpose:** Estimated mediation through PISA scores. **Visual forms:** Path diagram.
**Physical source pages:** 12.
[PNG](<07_Mechanisms_and_Heterogeneity/P141-Fig04-Effects%20of%20Three%20Teachings%20on%20economic%20growth%20directly%20and%20mediated%20via%20PISA%20scores.%20Notes.%20Standardized%20path%20coefficients%20(%CE%B2)%20with%20correlations%20(r)%20in....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=576>) · Original paper and literature note are not hosted.
### P144 - Fig. 3
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure 3: AI Diffusion Gap by Disadvantage Notes: This figure reports evidence of the AI diffusion gap using alternative measures of school disadvantage. The outcome is a standardized AI integration index that aggregates seven dimensions of school-level AI integration: written AI policy; stance toward student AI use; teacher AI adoption; principal AI use; expected future integration; AI tutoring availability; and AI professional development. Panels (a)-(c) plot binned scatterplots relating the index to free/reduced-price lunch (FRPL) shares, SEDA achievement, and log neighborhood median income, respectively. Panels (a) and (b) are restricted to public schools because FRPL and SEDA variables are unavailable for private schools; Panel (c) includes both public and private schools. Panel (d) reports standardized regression slope coefficients and 95 percent confidence intervals, where each disadvantage measure is normalized to mean zero and standard deviation one. All regressions are IPW-weighted and use heteroskedasticity-robust standard errors.
**Purpose:** AI diffusion differs by school disadvantage. **Visual forms:** Scatter, Forest/coefficient plot, Composite.
**Physical source pages:** 15.
[PNG](<07_Mechanisms_and_Heterogeneity/P144-Fig03-AI%20Diffusion%20Gap%20by%20Disadvantage%20Notes%20This%20figure%20reports%20evidence%20of%20the%20AI%20diffusion%20gap%20using%20alternative%20measures%20of%20school%20disadvantage.%20The%20outcome%20is%20a....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=577>) · Original paper and literature note are not hosted.
### P144 - Fig. 4
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure 4: AI Diffusion Gap by Sector Notes: This figure reports charter and private school gaps in the standardized AI integration index relative to traditional public schools. The AI integration index is the seven-dimension measure. Each bar reports the coefficient on a charter school or private school indicator from a separate regression of the standardized AI integration index on the sector indicator. Unadjusted bars (maroon) report unconditional differences. Adjusted bars (grey) additionally control for log neighborhood median income, neighborhood Black, Hispanic, and Asian shares, missingness indicators for each control, and state fixed effects. Charter school regressions are estimated on the public school sample (traditional public versus charter schools), while private school regressions are estimated on the non-charter sample (traditional public versus private schools). All regressions are IPW-weighted. Whiskers denote 95 percent confidence intervals based on heteroskedasticity-robust standard errors.
**Purpose:** AI diffusion differs by public, charter and private school sectors. **Visual forms:** Bar chart.
**Physical source pages:** 16.
[PNG](<07_Mechanisms_and_Heterogeneity/P144-Fig04-AI%20Diffusion%20Gap%20by%20Sector%20Notes%20This%20figure%20reports%20charter%20and%20private%20school%20gaps%20in%20the%20standardized%20AI%20integration%20index%20relative%20to%20traditional%20public....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=578>) · Original paper and literature note are not hosted.
### P144 - Fig. 5
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure 5: Factors that Predict AI Integration and Explain Diffusion Gaps Notes: This figure reports regression estimates assessing predictors of AI integration and the extent to which those predictors explain the diffusion gaps. The outcome is the standardized AI integration index. Panel (a) reports coefficients on school-level predictors from a single regression estimated on the traditional public school sample that controls for standardized FRPL share, demographic missingness indicators, and state fixed effects. The predictors are grouped into six categories: principal beliefs, leadership and discretion, budget flexibility and ESSER use, perceived barriers to adoption, cell-phone restrictions, and district/system capacity. The displayed coefficients in Panel (a) include whether the principal believes AI has improved student learning (Beliefs); whether the principal credits teacher champions for enabling AI adoption and has authority to purchase instructional software (Leadership); and whether the school has sufficient budget and flexibility to direct funds toward AI purchases (Budget). Panel (b) reports coefficients on each diffusion gap across sequential specifications that cumulatively add the predictor groups included in Panel (a). Disadvantage coefficients are estimated on the traditional public school sample; charter school coefficients are estimated on the public school sample (traditional public versus charter schools); private school coefficients are estimated on the non-charter sample (traditional public versus private schools). District/system capacity specifications are estimated only for the FRPL-gap analysis because district scale and fiscal capacity measures are not defined comparably for charter and private schools. All regressions are IPW-weighted. Whiskers denote 95 percent confidence intervals based on heteroskedasticity-robust standard errors.
**Purpose:** Predictors and sequential mechanism specifications explain diffusion gaps. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 18.
[PNG](<07_Mechanisms_and_Heterogeneity/P144-Fig05-Factors%20that%20Predict%20AI%20Integration%20and%20Explain%20Diffusion%20Gaps%20Notes%20This%20figure%20reports%20regression%20estimates%20assessing%20predictors%20of%20AI%20integration%20and%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=579>) · Original paper and literature note are not hosted.
### P144 - Fig. A.2
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure A.2: AI Integration and Beliefs about AI’s Impact on Student Learning Notes: This figure reports coefficients from linear probability models regressing principals’ positive student-learningimpact belief on the standardized AI integration index and each standardized index pillar. The outcome is an indicator equal to one when the principal answered “probably yes” or “definitely yes” to “Do you believe Generative AI tools have produced measurable improvements in student learning at your school?” The full index is the seven-pillar AI integration index defined in Appendix A.1; component rows use the corresponding standardized pillars. Unadjusted estimates include no controls or fixed effects. Adjusted estimates control for standardized FRPL disadvantage, school racial composition, missingness indicators, and state fixed effects. All regressions are estimated on the common traditional-public-school sample, use IPW weights, and report 95 percent confidence intervals based on heteroskedasticity-robust standard errors.
**Purpose:** AI integration related to beliefs about student learning. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 29.
[PNG](<07_Mechanisms_and_Heterogeneity/P144-FigA.2-AI%20Integration%20and%20Beliefs%20about%20AI%E2%80%99s%20Impact%20on%20Student%20Learning%20Notes%20This%20figure%20reports%20coefficients%20from%20linear%20probability%20models%20regressing%20principals%E2%80%99....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=580>) · Original paper and literature note are not hosted.
### P144 - Fig. B.2
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure B.2: Adoption and Professional Development by Sector Notes: This figure reports teacher AI adoption and AI professional development trends by school sector for 2022– 2025. Panel (a) reports the IPW-weighted share of schools for which principals reported that teachers were using Generative AI tools for teaching-related tasks on or before June of each year, separately for traditional public, charter, and private schools. Panel (b) reports the analogous share of schools for which principals reported that staff had received any professional development training on generative AI by the same June dates. The stars reported above charter and private points correspond to the significance level of unadjusted differences relative to traditional public schools in a given year: ∗p < 0.10, ∗∗ p < 0.05,∗∗∗ p < 0.01. All estimates are IPW-weighted.
**Purpose:** Adoption trends compared across school sectors. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 33.
[PNG](<07_Mechanisms_and_Heterogeneity/P144-FigB.2-Adoption%20and%20Professional%20Development%20by%20Sector%20Notes%20This%20figure%20reports%20teacher%20AI%20adoption%20and%20AI%20professional%20development%20trends%20by%20school%20sector%20for%202022....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=581>) · Original paper and literature note are not hosted.
### P144 - Fig. B.3
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure B.3: ESSER Funds and AI Adoption Notes: Panel (a) reports the IPW-weighted share of schools whose principals reported that their school had used federal pandemic-relief funds (e.g., ESSER) for AI-related purchases, separately for traditional public, charter, and private schools. Principals could answer yes, no, or not sure; the plotted indicator codes yes as reported ESSER use and no or not sure as not reported ESSER use. Panel (b) plots the IPW-weighted share of schools for which principals reported that teachers were using Generative AI tools for teaching-related tasks on or before June of each year, 2022–2025, separately by this reported ESSER-use indicator. The vertical axes show percentages.
**Purpose:** Adoption differs by funding use and school sector. **Visual forms:** Bar chart, Line/profile plot, Composite.
**Physical source pages:** 33.
[PNG](<07_Mechanisms_and_Heterogeneity/P144-FigB.3-ESSER%20Funds%20and%20AI%20Adoption%20Notes%20Panel%20(a)%20reports%20the%20IPW-weighted%20share%20of%20schools%20whose%20principals%20reported%20that%20their%20school%20had%20used%20federal....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=582>) · Original paper and literature note are not hosted.
### P144 - Fig. C.1
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure C.1: Component-Level FRPL Gaps Notes: Each point is the coefficient on FRPL share from a separate regression of the indicated AI index component on FRPL, with state fixed effects, IPW weights, and missingness controls. Components are: Policy, written districtor school-level generative AI policy/topic depth, coded 0 for no written policy or draft only, 1 for a written policy, and 2 for a written policy covering at least three of five listed topics (student data privacy, academic honesty/plagiarism, special-education accommodations, accessibility, and content filtering/safety); Student policy, whether the current stance communicated to students is “Actively encouraged for learning, including studying and homework assistance, with clear rules”; Teacher adopt, the mean of four indicators for teacher use of generative AI tools for teaching-related tasks on or before June 2022, June 2023, June 2024, and June 2025; Principal use, the mean of four indicators for principal use to offload administrative busywork, streamline data analysis/reporting, design professional development, and prepare meetings/presentations; Future plans, expected “Extensive–most instruction” or “Embedded–routine in all classes” integration three years ahead; Tutoring, a current license or contract for any AI-powered tutoring platform; and PD, staff generative AI professional development by June 2025, with an additional point when at least 25% of instructional staff completed at least two hours of such professional development. Error bars are 95% CIs.
**Purpose:** Component-level economic-disadvantage heterogeneity. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 42.
[PNG](<07_Mechanisms_and_Heterogeneity/P144-FigC.1-Component-Level%20FRPL%20Gaps%20Notes%20Each%20point%20is%20the%20coefficient%20on%20FRPL%20share%20from%20a%20separate%20regression%20of%20the%20indicated%20AI%20index%20component%20on%20FRPL,%20with%20state....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=583>) · Original paper and literature note are not hosted.
### P144 - Fig. D.1
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure D.1: Factors that May Explain the Diffusion Gap Notes: This figure reports seven principal-reported mechanism variables by school disadvantage and sector. The left column splits public schools at the median FRPL share, with above-median schools classified as high disadvantage; the right column compares traditional public, charter, and private schools. Variables cover belief, leadership, budget, and professional-development channels. Belief measures indicate whether principals report measurable student-learning improvements from generative AI and whether they agree that generative AI is widening achievement gaps. Leadership measures indicate principal authority to purchase instructional software and teacher champions as the top enabler of current AI efforts. Budget measures indicate a discretionary technology budget of at least $5,000 and whether redirecting funds toward AI is easy or very easy. The professional-development measure indicates agreement that AI-related PD improved lesson-planning efficiency, conditional on the school having received AI PD. All shares are IPW-weighted.
**Purpose:** Proposed mechanisms compared across disadvantage and school sector. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 49.
[PNG](<07_Mechanisms_and_Heterogeneity/P144-FigD.1-Factors%20that%20May%20Explain%20the%20Diffusion%20Gap%20Notes%20This%20figure%20reports%20seven%20principal-reported%20mechanism%20variables%20by%20school%20disadvantage%20and%20sector.%20The%20left....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=584>) · Original paper and literature note are not hosted.
### P144 - Fig. D.2
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure D.2: Cell-Phone Restrictions and Teacher AI Adoption Notes: This figure reports coefficients from separate regressions of the standardized teacher-adoption pillar on each recoded cell-phone restriction measure. Cell-phone measures are coded from public district- or school-level policy sources, not the principal survey; the 0–5 index codes 0 for no restriction found, 1 for weak or silenced-only policies, 2 for classroom-only restrictions, 3 for full-day use restrictions, 4 for full-day storage restrictions, and 5 for possession bans or locked-storage requirements. The teacher-adoption pillar is the mean of four yes/no principal reports of whether teachers at their school were using Generative AI tools for teaching-related tasks on or before June 2022, June 2023, June 2024, and June 2025, standardized using IPW weights. The 0–5 restriction index is standardized using IPW weights, so its coefficient reports the association with a one standard deviation increase in policy restrictiveness. The remaining measures are binary indicators for school-day restrictions, storage-or-stronger restrictions, and possession or locked-storage restrictions. All regressions control for school sector, log neighborhood median household income, school racial composition, school level, log enrollment, missingness indicators, and state fixed effects. Regressions are IPW-weighted and use heteroskedasticity-robust standard errors to construct the reported 95 percent confidence intervals.
**Purpose:** Phone restrictions related to teacher AI adoption. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 50.
[PNG](<07_Mechanisms_and_Heterogeneity/P144-FigD.2-Cell-Phone%20Restrictions%20and%20Teacher%20AI%20Adoption%20Notes%20This%20figure%20reports%20coefficients%20from%20separate%20regressions%20of%20the%20standardized%20teacher-adoption%20pillar%20on....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=585>) · Original paper and literature note are not hosted.
### P144 - Fig. E.4
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure E.4: Adoption by Principal Characteristics Notes: This figure reports IPW-weighted shares of schools where principals answered yes that teachers at their school were using Generative AI tools for teaching-related tasks on or before June of each year, by principal gender, race/ethnicity, and years served as principal at this school. Panel (a) uses the gender item; Panel (b) uses the race/ethnicity item and reports White, Black, and Hispanic or Latino groups; Panel (c) groups reported years served as principal at this school into 1–3, 4–7, and 8+ years. Stars denote significant pairwise differences relative to the reference group (∗p < 0.10, ∗∗ p < 0.05,∗∗∗ p < 0.01). Reference groups: male, White, and 1–3 years.
**Purpose:** Adoption patterns differ by principal characteristics. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 54.
[PNG](<07_Mechanisms_and_Heterogeneity/P144-FigE.4-Adoption%20by%20Principal%20Characteristics%20Notes%20This%20figure%20reports%20IPW-weighted%20shares%20of%20schools%20where%20principals%20answered%20yes%20that%20teachers%20at%20their%20school%20were....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=586>) · Original paper and literature note are not hosted.
### P149 - Fig. 4
**Paper:** A systematic review and meta-analysis of the evidence on learning during the COVID-19 pandemic
**Original caption:** Fig. 4 | Estimates of COVID-19 learning deficits (n = 291), by date of measurement. The horizontal axis displays the date on which learning progress was measured. The vertical axis displays estimated learning deficits, expressed in standard deviation (s.d.) using Cohen’s d. The colour of the circles reflects the respective country, the size of the circles indicates the sample size for a given estimate and the line displays a linear trend with a 95% CI. The trend line is estimated as a linear regression using ordinary least squares, with standard errors clustered at the study level (n = 42 clusters). β_months = −0.00, t(41) = −7.30, two-tailed P = 0.097, 95% CI −0.01 to 0.00.
**Purpose:** Examines learning-deficit variation over pandemic time. **Visual forms:** Bubble scatterplot.
**Physical source pages:** 5.
[PNG](<07_Mechanisms_and_Heterogeneity/P149-Fig04-Estimates%20of%20COVID-19%20learning%20deficits%20(n%20=%20291),%20by%20date%20of%20measurement.%20The%20horizontal%20axis%20displays%20the%20date%20on%20which%20learning%20progress%20was%20measured.%20The....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=587>) · Original paper and literature note are not hosted.
### P149 - Fig. 5
**Paper:** A systematic review and meta-analysis of the evidence on learning during the COVID-19 pandemic
**Original caption:** Fig. 5 | Harvest plot summarizing the evidence on changes in educational inequality between students from different socio-economic backgrounds during the pandemic. Each circle/square refers to one estimate of over-time change in inequality in maths/reading performance (n = 211). Estimates that find a decrease/no change/increase in inequality are grouped on the left/middle/ right. Within these categories, estimates are ordered horizontally by school grade. The shading indicates when in the pandemic a given measure was taken.
**Purpose:** Summarizes changes in educational inequality across socioeconomic backgrounds. **Visual forms:** Harvest plot.
**Physical source pages:** 5.
[PNG](<07_Mechanisms_and_Heterogeneity/P149-Fig05-Harvest%20plot%20summarizing%20the%20evidence%20on%20changes%20in%20educational%20inequality%20between%20students%20from%20different%20socio-economic%20backgrounds%20during%20the%20pandemic.%20Each....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=588>) · Original paper and literature note are not hosted.
### P149 - Fig. 6
**Paper:** A systematic review and meta-analysis of the evidence on learning during the COVID-19 pandemic
**Original caption:** Fig. 6 | Variation in estimates of COVID-19 learning deficits (n = 291) across different characteristics. Each plot shows the distribution of COVID-19 learning deficit estimates for the respective subgroup, with the box marking the interquartile range and the white circle denoting the median. Whiskers mark upper and lower adjacent values: the furthest observation within 1.5 interquartile range of either side of the box. a, Learning subject (reading versus maths). Median: reading −0.09, maths −0.18. Interquartile range: reading −0.15 to −0.02, maths −0.23 to −0.09. b, Level of education (primary versus secondary). Median: primary −0.12, secondary −0.12. Interquartile range: primary −0.19 to −0.05, secondary −0.21 to −0.06. c, Country income level (high versus middle). Median: high −0.12, middle −0.37. Interquartile range: high −0.20 to −0.05, middle −0.65 to −0.30.
**Purpose:** Compares learning deficits by subject, education and country-income groups. **Visual forms:** Composite, Violin plot.
**Physical source pages:** 6.
[PNG](<07_Mechanisms_and_Heterogeneity/P149-Fig06-Variation%20in%20estimates%20of%20COVID-19%20learning%20deficits%20(n%20=%20291)%20across%20different%20characteristics.%20Each%20plot%20shows%20the%20distribution%20of%20COVID-19%20learning%20deficit....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=589>) · Original paper and literature note are not hosted.
### P153 - Fig. 2
**Paper:** Do university-educated families lose their edge as education expands? The withering performance and advantage of their children
**Original caption:** Figure 2. Average Marginal Effect of the longitudinal component of the proportion of university-educated parents on children’s academic achievement by parental education. Note: *** p-value ≤ 0.001; ** p-value ≤ 0.05; * p-value ≤ 0.10. Standard errors between brackets. All models control for country-by-wave GDP per capita, population, and share of the student population with a migrant background, and include wave-fixed effects. All these variables are interacted with parental education. Source: Authors’ elaboration from PISA 2003 to 2022.
**Purpose:** Association of educational expansion differs by parental education. **Visual forms:** Bar chart.
**Physical source pages:** 23.
[PNG](<07_Mechanisms_and_Heterogeneity/P153-Fig02-Average%20Marginal%20Effect%20of%20the%20longitudinal%20component%20of%20the%20proportion%20of%20university-educated%20parents%20on%20children%E2%80%99s%20academic%20achievement%20by%20parental%20education....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=590>) · Original paper and literature note are not hosted.
### P153 - Fig. 3
**Paper:** Do university-educated families lose their edge as education expands? The withering performance and advantage of their children
**Original caption:** Figure 3. Mediation analysis for the effect of the longitudinal component of the expansion on the achievement of children from university-educated families. Note: *** p-value ≤ 0.001; ** p-value ≤ 0.05; * p-value ≤ 0.10. Standard errors between brackets. All models control for country-by-wave GDP per capita, population, and share of the student population with a migrant background, and include wave-fixed effects. All these variables are interacted with parental education. Model 1 is the baseline model. Model 2 includes the number of overqualified universityeducated parents. Model 3 includes the number of books at home. Model 4 includes both mediators. In grey, the percentage of mediation comparing with the baseline model. Source: Authors’ elaboration from PISA 2003 to 2022.
**Purpose:** Mediation by parental occupational mismatch and home books. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 24.
[PNG](<07_Mechanisms_and_Heterogeneity/P153-Fig03-Mediation%20analysis%20for%20the%20effect%20of%20the%20longitudinal%20component%20of%20the%20expansion%20on%20the%20achievement%20of%20children%20from%20university-educated%20families.%20Note%203....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=591>) · Original paper and literature note are not hosted.
### P155 - Fig. 2
**Paper:** Educational participation, cultural resources, and social inequality: Have social inequalities in reading literacy increased or decreased between 2000 and 2022?
**Original caption:** Figure 2. Educational participation by social background (ISEI-quartile) between 2000 and 2022 (in percent). Source. PISA 2000–2022 (imputed and clustered data; own calculation). PISA: Programme for International Student Assessment; ISEI: International Socio-Economic Index of Occupational Status.
**Purpose:** SES-quartile contrasts test inequality in access to school tracks. **Visual forms:** Bar chart.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P155-Fig02-Educational%20participation%20by%20social%20background%20(ISEI-quartile)%20between%202000%20and%202022%20(in%20percent).%20Source.%20PISA%202000-2022%20(imputed%20and%20clustered%20data;%20own....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=592>) · Original paper and literature note are not hosted.
### P155 - Fig. 3
**Paper:** Educational participation, cultural resources, and social inequality: Have social inequalities in reading literacy increased or decreased between 2000 and 2022?
**Original caption:** Figure 3. Number of books by social background (ISEI-quartile) between 2000 and 2022 (in percent). Source. PISA 2000–2022 (imputed and clustered data; own calculation). PISA: Programme for International Student Assessment; ISEI: International Socio-Economic Index of Occupational Status.
**Purpose:** SES-quartile contrasts test inequality in family cultural resources. **Visual forms:** Bar chart.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P155-Fig03-Number%20of%20books%20by%20social%20background%20(ISEI-quartile)%20between%202000%20and%202022%20(in%20percent).%20Source.%20PISA%202000-2022%20(imputed%20and%20clustered%20data;%20own%20calculation)....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=593>) · Original paper and literature note are not hosted.
### P155 - Fig. 4
**Paper:** Educational participation, cultural resources, and social inequality: Have social inequalities in reading literacy increased or decreased between 2000 and 2022?
**Original caption:** Figure 4. Correlation between standardized socio-economic status (ISEI) and educational participation and cultural resources between 2000 and 2022 (R²). Source. PISA 2000–2022 (imputed and clustered data; own calculation). Notes. Full information on the regression models can be found in Tables A3 and A4 in the appendix. PISA: Programme for International Student Assessment; ISEI: International Socio-Economic Index of Occupational Status.
**Purpose:** Tests changing links of social background to school participation and cultural resources. **Visual forms:** Scatter.
**Physical source pages:** 8.
[PNG](<07_Mechanisms_and_Heterogeneity/P155-Fig04-Correlation%20between%20standardized%20socio-economic%20status%20(ISEI)%20and%20educational%20participation%20and%20cultural%20resources%20between%202000%20and%202022%20(R2).%20Source.%20PISA....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=594>) · Original paper and literature note are not hosted.
### P155 - Fig. 5
**Paper:** Educational participation, cultural resources, and social inequality: Have social inequalities in reading literacy increased or decreased between 2000 and 2022?
**Original caption:** Figure 5. Correlation between reading skills and educational participation and cultural resources between 2000 and 2022 (R2 ). Source. PISA 2000–2022 (imputed and clustered data; own calculation). Notes. Full information on the regression models can be found in Tables A5 and A6 in the appendix. PISA: Programme for International Student Assessment.
**Purpose:** Compares the changing explanatory roles of school tracks and cultural resources. **Visual forms:** Scatter.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P155-Fig05-Correlation%20between%20reading%20skills%20and%20educational%20participation%20and%20cultural%20resources%20between%202000%20and%202022%20(R2%20).%20Source.%20PISA%202000-2022%20(imputed%20and%20clustered....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=595>) · Original paper and literature note are not hosted.
### P155 - Fig. 7
**Paper:** Educational participation, cultural resources, and social inequality: Have social inequalities in reading literacy increased or decreased between 2000 and 2022?
**Original caption:** Figure 7. Results of KHB-decompositions. Share of social differences in reading (2000–2022) and explanatory part of educational participation and cultural conditions. Source. PISA 2000–2022 (imputed and clustered data; own calculation). Notes. The lengths of bars show the regression coefficients of parental ISEI on students’ reading skills in the respective years (total ISEI effect). The coloured areas represent the explanatory contribution of the corresponding variables to the skill differences (see Table A7 in appendix). PISA: Programme for International Student Assessment; ISEI: International Socio-Economic Index of Occupational Status.
**Purpose:** KHB decomposition attributes the reading gradient to school, family and control-variable components. **Visual forms:** Bar chart.
**Physical source pages:** 11.
[PNG](<07_Mechanisms_and_Heterogeneity/P155-Fig07-Results%20of%20KHB-decompositions.%20Share%20of%20social%20differences%20in%20reading%20(2000-2022)%20and%20explanatory%20part%20of%20educational%20participation%20and%20cultural%20conditions....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=596>) · Original paper and literature note are not hosted.
### P158 - Table 5
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Table 5 Studies included in the meta-analysis. p < 0.05 = *; p < 0.01 = **; p < 0.001 = ***
**Purpose:** Multi-level study coding compares candidate moderators and learning-outcome evidence. **Visual forms:** Complex table, Structured evidence comparison.
**Physical source pages:** 86, 87, 88.
[PNG](<07_Mechanisms_and_Heterogeneity/P158-Table05-Studies%20included%20in%20the%20meta-analysis.%20p%20less%20than%200.05%20=%201%20asterisk%20;%20p%20less%20than%200.01%20=%202%20asterisks%20;%20p%20less%20than%200.001%20=%203%20asterisks.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=597>) · Original paper and literature note are not hosted.
### P158 - Fig. 7
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 7 Posterior model-averaged distribution of the effect size estimate for the moderators "learning outcome" based on the RoBMA. Please note that the height of the arrow corresponds to the second y-axis on the right indicating the point probability mass allocated to the Null Hypothesis H0 of the level "skills". The lines represent the density distribution for knowledge (K) and skills (S)
**Purpose:** Moderator-specific estimates by learning outcome. **Visual forms:** Density plot.
**Physical source pages:** 27.
[PNG](<07_Mechanisms_and_Heterogeneity/P158-Fig07-Posterior%20model-averaged%20distribution%20of%20the%20effect%20size%20estimate%20for%20the%20moderators%20learning%20outcome%20based%20on%20the%20RoBMA.%20Please%20note%20that%20the%20height%20of%20the%20arrow....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=598>) · Original paper and literature note are not hosted.
### P158 - Fig. 8
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 8 Posterior model-averaged distribution of the effect size estimate for the moderators "ISAR level" based on the RoBMA. Please note that the height of the arrow corresponds to the second y-axis on the right indicating the point probability mass allocated to the Null Hypothesis H0 of the level "Substitution". The lines represent the density distribution for Substitution (S), Augmentation (A) and Redefinition (R)
**Purpose:** Moderator-specific estimates by ISAR level. **Visual forms:** Density plot.
**Physical source pages:** 27.
[PNG](<07_Mechanisms_and_Heterogeneity/P158-Fig08-Posterior%20model-averaged%20distribution%20of%20the%20effect%20size%20estimate%20for%20the%20moderators%20ISAR%20level%20based%20on%20the%20RoBMA.%20Please%20note%20that%20the%20height%20of%20the%20arrow....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=599>) · Original paper and literature note are not hosted.
### P158 - Fig. 9
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 9 Illustration of the robust effect of ISAR levels and learning outcome when implementing GAI in STEM education
**Purpose:** Summarizes moderator patterns for engagement level and learning outcome. **Visual forms:** Conceptual evidence synthesis.
**Physical source pages:** 29.
[PNG](<07_Mechanisms_and_Heterogeneity/P158-Fig09-Illustration%20of%20the%20robust%20effect%20of%20ISAR%20levels%20and%20learning%20outcome%20when%20implementing%20GAI%20in%20STEM%20education.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=600>) · Original paper and literature note are not hosted.
### P159 - Fig. A10
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure A10: Types of Generative AI Use by Augmentation, Automation, and Mixed Users. Notes: This figure replicates Figure 3 separately for augmentation, automation, and mixed users, as classified in Section 5.3. See notes to that figure for details.
**Purpose:** Compares reported and logged AI-use categories across augmentation, automation, and mixed users. **Visual forms:** Composite.
**Physical source pages:** 52.
[PNG](<07_Mechanisms_and_Heterogeneity/P159-FigA10-Types%20of%20Generative%20AI%20Use%20by%20Augmentation,%20Automation,%20and%20Mixed%20Users.%20Notes%20This%20figure%20replicates%20Figure%203%20separately%20for%20augmentation,%20automation,%20and%20mixed....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=601>) · Original paper and literature note are not hosted.
### P161 - Fig. 4
**Paper:** Extramural ICT factors impact adolescents’ academic performance and well-being differently: Types of self-regulated learners also matter
**Original caption:** Fig. 4 Mean comparisons of self-regulated learning across profiles. Notes. UNDREM = understanding and remembering; METASUM = summarizing; METASPAM = assessing credibility; JOY = reading enjoyment; SCOMP = reading self-concept. Positive values indicate above the means across OECD economies while negative values are below OECD means
**Purpose:** Learner-profile indicator patterns compared across Hong Kong and the USA. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 20.
[PNG](<07_Mechanisms_and_Heterogeneity/P161-Fig04-Mean%20comparisons%20of%20self-regulated%20learning%20across%20profiles.%20Notes.%20UNDREM%20=%20understanding%20and%20remembering;%20METASUM%20=%20summarizing;%20METASPAM%20=%20assessing....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=602>) · Original paper and literature note are not hosted.
### P170 - Fig. 1
**Paper:** Introduction: Administrative Burden as a Mechanism of Inequality in Policy Implementation
**Original caption:** Figure 1. The Enduring Racialized Effect of Burdens in Voting, Louisiana Registration Rates, 1878–2000 Source: Keele, Cubbison, and White 2021. Reprinted with permission.
**Purpose:** Historical racial disparity illustrates administrative-burden mechanisms. **Visual forms:** Line/profile plot.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P170-Fig01-The%20Enduring%20Racialized%20Effect%20of%20Burdens%20in%20Voting,%20Louisiana%20Registration%20Rates,%201878-2000%20Source%20Keele,%20Cubbison,%20and%20White%202021.%20Reprinted%20with%20permission.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=603>) · Original paper and literature note are not hosted.
### P170 - Fig. 2
**Paper:** Introduction: Administrative Burden as a Mechanism of Inequality in Policy Implementation
**Original caption:** Figure 2. Observed and Counterfactual Replacement Rates If Families Experiencing Joblessness Received All Income Transfers for Which They Are Eligible (No “Administrative Burden” Effect)
**Purpose:** Observed and counterfactual benefit replacement compared across race. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 8.
[PNG](<07_Mechanisms_and_Heterogeneity/P170-Fig02-Observed%20and%20Counterfactual%20Replacement%20Rates%20If%20Families%20Experiencing%20Joblessness%20Received%20All%20Income%20Transfers%20for%20Which%20They%20Are%20Eligible%20(No%20%E2%80%9CAdministrative....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=604>) · Original paper and literature note are not hosted.
### P174 - Fig. 4
**Paper:** Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach
**Original caption:** FIGURE 4.—1988 density of men’s real log wages ($1979) adjusted for the indicated factors.
**Purpose:** Sequential components explain institutional and compositional mechanisms in male wage-density change. **Visual forms:** Density, Composite.
**Physical source pages:** 25.
[PNG](<07_Mechanisms_and_Heterogeneity/P174-Fig04-1988%20density%20of%20men%E2%80%99s%20real%20log%20wages%20($1979)%20adjusted%20for%20the%20indicated%20factors.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=605>) · Original paper and literature note are not hosted.
### P174 - Fig. 5
**Paper:** Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach
**Original caption:** FIGURE 5.—1988 density of women’s real log wages ($1979) adjusted for the indicated factors.
**Purpose:** Sequential components explain institutional and compositional mechanisms in female wage-density change. **Visual forms:** Density, Composite.
**Physical source pages:** 26.
[PNG](<07_Mechanisms_and_Heterogeneity/P174-Fig05-1988%20density%20of%20women%E2%80%99s%20real%20log%20wages%20($1979)%20adjusted%20for%20the%20indicated%20factors.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=606>) · Original paper and literature note are not hosted.
### P174 - Fig. 6
**Paper:** Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach
**Original caption:** FIGURE 6.—Smoothed differences between the 1988 density adjusted for indicated factors and the 1979 density for men.
**Purpose:** Decomposed density differences show male institutional/composition contributions. **Visual forms:** Density, Composite.
**Physical source pages:** 28.
[PNG](<07_Mechanisms_and_Heterogeneity/P174-Fig06-Smoothed%20differences%20between%20the%201988%20density%20adjusted%20for%20indicated%20factors%20and%20the%201979%20density%20for%20men.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=607>) · Original paper and literature note are not hosted.
### P174 - Fig. 7
**Paper:** Labor Market Institutions and the Distribution of Wages, 1973-1992: A Semiparametric Approach
**Original caption:** FIGURE 7.—Smoothed differences between the 1988 density adjusted for indicated factors and the 1979 density for women.
**Purpose:** Decomposed density differences show female institutional/composition contributions. **Visual forms:** Density, Composite.
**Physical source pages:** 29.
[PNG](<07_Mechanisms_and_Heterogeneity/P174-Fig07-Smoothed%20differences%20between%20the%201988%20density%20adjusted%20for%20indicated%20factors%20and%20the%201979%20density%20for%20women.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=608>) · Original paper and literature note are not hosted.
### P175 - Fig. 3
**Paper:** Machine learning and phone data can improve targeting of humanitarian aid
**Original caption:** Fig. 3 | Fairness of targeting for different demographic subgroups. a, b, Distributions of differences between ranking according to predicted wealth from the phone-based approach and ranking according to true wealth (using the 2018–2019 field survey; n = 4,171), dissagregated by gender (a) and ethnicity (b). Boxes show the 25th and 75th percentiles, whiskers show the minimum and maximum, and the centre line shows the median of the distribution. Left-skewed bars indicate groups that are consistently under-ranked; right-skewed bars indicate groups that are consistently over-ranked. c, d, Evaluation of demographic parity across subgroups by comparing the proportion of a subgroup targeted under counterfactual approaches to the proportion of the subgroup that falls into the poorest 29% of the population (using the 2018–2019 field survey; n = 4,171), disaggregated by gender (c) and ethnicity (d). Bubbles show the percentage point difference between the proportion of the subgroup that is targeted and the proportion that is poor according to ground-truth data. Large red bubbles indicate groups that are over-targeted; large blue bubbles indicate groups that are under-targeted.
**Purpose:** Gender/ethnic heterogeneity in ranking errors and demographic parity. **Visual forms:** Boxplot, Bubble matrix, Composite.
**Physical source pages:** 5.
[PNG](<07_Mechanisms_and_Heterogeneity/P175-Fig03-Fairness%20of%20targeting%20for%20different%20demographic%20subgroups.%20a,%20b,%20Distributions%20of%20differences%20between%20ranking%20according%20to%20predicted%20wealth%20from%20the%20phone-based....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=609>) · Original paper and literature note are not hosted.
### P175 - Fig. Extended Data 2
**Paper:** Machine learning and phone data can improve targeting of humanitarian aid
**Original caption:** Extended Data Fig. 2 | Fairness with normalized rank residuals. Boxplots showing distributions of normalized rank residuals (see Methods, ‘Fairness’) aggregated by subgroup, using the 2018–2019 field survey dataset (N = 4,171). Boxes show the 25th and 75th percentiles, and the centre line shows the median of the distribution. Left-shifted boxes indicate groups that are consistently under-ranked by a given targeting mechanism, right-shifted boxes indicate groups that are consistently over-ranked by a given targeting mechanism.
**Purpose:** Expanded subgroup fairness comparison using normalized rank residuals. **Visual forms:** Boxplot, Composite.
**Physical source pages:** 19.
[PNG](<07_Mechanisms_and_Heterogeneity/P175-FigExtended%20Data%202-Extended%20Data%20Fig.%202%20Fairness%20with%20normalized%20rank%20residuals.%20Boxplots%20showing%20distributions%20of%20normalized%20rank%20residuals%20(see%20Methods,%20%E2%80%98Fairness%E2%80%99)....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=610>) · Original paper and literature note are not hosted.
### P175 - Fig. Extended Data 3
**Paper:** Machine learning and phone data can improve targeting of humanitarian aid
**Original caption:** Extended Data Fig. 3 | Fairness with demographic parity. We evaluate demographic parity across subgroups by comparing the proportion of a subgroup targeted under counterfactual approaches to the proportion of the subgroup that falls into the poorest 29% of the population (using data from the 2018–2019 field survey matched to CDR, N = 4,171). Bubbles show the percentage point difference between the proportion of the subgroup that is targeted and the proportion that is poor according to ground-truth data. Large red bubbles indicate groups that are over-targeted; large blue bubbles indicate groups that are under-targeted.
**Purpose:** Expanded demographic-parity comparisons across eight subgroup dimensions. **Visual forms:** Bubble matrix, Composite.
**Physical source pages:** 20.
[PNG](<07_Mechanisms_and_Heterogeneity/P175-FigExtended%20Data%203-Extended%20Data%20Fig.%203%20Fairness%20with%20demographic%20parity.%20We%20evaluate%20demographic%20parity%20across%20subgroups%20by%20comparing%20the%20proportion%20of%20a%20subgroup....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=611>) · Original paper and literature note are not hosted.
### P182 - Fig. 5
**Paper:** Patterns of item nonresponse behaviour to survey questionnaires are systematic and associated with genetic loci
**Original caption:** Fig. 5 | Genetic correlations between the non-response factors and other heritable traits. a,b, Genetic correlations between our PNA (a) and IDK (b) factors and 655 other traits. The values on the x axis represent the point estimate of the genetic correlation. The values on the y axis represent −log10 of the P value of the associated test statistic. Only traits with a genetic correlation (in absolute value) >0.50 and with P < 10−50 are labelled. c, Relationship between genetic correlation point estimates for other traits with PNA and IDK. Point estimates for PNA are shown on the x axis and those for IDK on the y axis. In a and b, marker colour reflects the category of the trait. In c, colour indicates the significance level (that is, not significant: P > 0.05, nominally significant: P < 0.05, significant after multiple-hypothesis Bonferroni correction for 654 traits tested: P < 7.65 × 10−5) for the difference between PNA and IDK correlations with a given trait. All statistical tests for a, b and c are two-sided.
**Purpose:** Compares PNA and IDK genetic associations across a broad trait set and tests differences between them. **Visual forms:** Composite, Scatter, Density/histogram/boxplot.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P182-Fig05-Genetic%20correlations%20between%20the%20non-response%20factors%20and%20other%20heritable%20traits.%20a,b,%20Genetic%20correlations%20between%20our%20PNA%20(a)%20and%20IDK%20(b)%20factors%20and%20655%20other....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=612>) · Original paper and literature note are not hosted.
### P188 - Table 3
**Paper:** Socioeconomic status and student learning: insights from an umbrella review
**Original caption:** Table 3 SES/Demographics/neighborhoods, SES as moderator, and moderators of SES effects
**Purpose:** Cross-review comparison of intersecting demographics, SES moderation and contextual effect heterogeneity. **Visual forms:** Complex table.
**Physical source pages:** 36, 37, 38, 39, 40, 41, 42, 43, 44, 45.
[PNG part 1](<07_Mechanisms_and_Heterogeneity/P188-Table03-SES%20Demographics%20neighborhoods,%20SES%20as%20moderator,%20and%20moderators%20of%20SES%20effects-Part01.png>) · [PNG part 2](<07_Mechanisms_and_Heterogeneity/P188-Table03-SES%20Demographics%20neighborhoods,%20SES%20as%20moderator,%20and%20moderators%20of%20SES%20effects-Part02.png>) · [PNG part 3](<07_Mechanisms_and_Heterogeneity/P188-Table03-SES%20Demographics%20neighborhoods,%20SES%20as%20moderator,%20and%20moderators%20of%20SES%20effects-Part03.png>) · [PNG part 4](<07_Mechanisms_and_Heterogeneity/P188-Table03-SES%20Demographics%20neighborhoods,%20SES%20as%20moderator,%20and%20moderators%20of%20SES%20effects-Part04.png>) · [PNG part 5](<07_Mechanisms_and_Heterogeneity/P188-Table03-SES%20Demographics%20neighborhoods,%20SES%20as%20moderator,%20and%20moderators%20of%20SES%20effects-Part05.png>) · [PNG part 6](<07_Mechanisms_and_Heterogeneity/P188-Table03-SES%20Demographics%20neighborhoods,%20SES%20as%20moderator,%20and%20moderators%20of%20SES%20effects-Part06.png>) · [PNG part 7](<07_Mechanisms_and_Heterogeneity/P188-Table03-SES%20Demographics%20neighborhoods,%20SES%20as%20moderator,%20and%20moderators%20of%20SES%20effects-Part07.png>) · [PNG part 8](<07_Mechanisms_and_Heterogeneity/P188-Table03-SES%20Demographics%20neighborhoods,%20SES%20as%20moderator,%20and%20moderators%20of%20SES%20effects-Part08.png>) · [PNG part 9](<07_Mechanisms_and_Heterogeneity/P188-Table03-SES%20Demographics%20neighborhoods,%20SES%20as%20moderator,%20and%20moderators%20of%20SES%20effects-Part09.png>) · [PNG part 10](<07_Mechanisms_and_Heterogeneity/P188-Table03-SES%20Demographics%20neighborhoods,%20SES%20as%20moderator,%20and%20moderators%20of%20SES%20effects-Part10.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=613>) · Original paper and literature note are not hosted.
### P191 - Fig. 5
**Paper:** Substitute or Complement? Generative AI Use and Student Achievement: Evidence from PISA 2025
**Original caption:** Figure 5: System heterogeneity in the AI gradient. Notes: PISA 2025, 35 OECD countries with at least 1,500 usable students. Panel (a) gives each country’s within-school level slope with its 95 percent interval, sorted. In panels (b) and (c) marker area is proportional to the precision of the country’s first-step estimate and the fitted line is weighted by that precision.
**Purpose:** Compares country-level slopes and tests how the AI-use and teacher-support context relates to heterogeneity. **Visual forms:** Composite, Forest/coefficient plot, Scatter.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P191-Fig05-System%20heterogeneity%20in%20the%20AI%20gradient.%20Notes%20PISA%202025,%2035%20OECD%20countries%20with%20at%20least%201,500%20usable%20students.%20Panel%20(a)%20gives%20each%20country%E2%80%99s%20within-school....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=623>) · Original paper and literature note are not hosted.
### P205 - Fig. 1
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** Figure 1 Comparing process maps of the revising stage between AI and CL groups
**Purpose:** Additional AI/checklist process comparison; original appendix numbering remains separate from main Figure 1. **Visual forms:** Process network.
**Physical source pages:** 36.
[PNG](<07_Mechanisms_and_Heterogeneity/P205-Fig01-Comparing%20process%20maps%20of%20the%20revising%20stage%20between%20AI%20and%20CL%20groups.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=915>) · Original paper and literature note are not hosted.
### P205 - Fig. 3
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** FIGURE 3 Comparative results of SRL process frequencies among four groups in two learning stages. Note: Four groups are Control group (CN), ChatGPT group (AI), human expert group (HE) and checklist feedback tools group (CL); Seven processes are Orientation process (MC.O), Planning process (MC.P), Monitoring process (MC.M), Evaluation process (MC.E), Reading process (LC.FR), Elaboration and Organisation process (HC.EO), and Other process (learners interacting with various agents). The Kruskal-Wallis test was performed for each stage between the treatment groups and the control group. Statistical significance levels are denoted as follows: **** for p<0.0001, *** for p<0.001, ** for p<0.01, * for p<0.05, and ns for “not significant”.
**Purpose:** SRL-process comparisons investigate how support changes regulation; both stages, all process panels and complete explanatory note retained. **Visual forms:** Boxplot, Composite.
**Physical source pages:** 14.
[PNG](<07_Mechanisms_and_Heterogeneity/P205-Fig03-Comparative%20results%20of%20SRL%20process%20frequencies%20among%20four%20groups%20in%20two%20learning%20stages.%20Note%20Four%20groups%20are%20Control%20group%20(CN),%20ChatGPT%20group%20(AI),%20human%20expert....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=916>) · Original paper and literature note are not hosted.
### P205 - Fig. 4
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** FIGURE 4 Comparative results of SRL process maps for AI/CN and AI/HE in the revising stage
**Purpose:** Complete original two-panel process-map comparison; source caption and both panel subtitles retained. **Visual forms:** Process network, Composite.
**Physical source pages:** 15.
[PNG](<07_Mechanisms_and_Heterogeneity/P205-Fig04-Comparative%20results%20of%20SRL%20process%20maps%20for%20AI%20CN%20and%20AI%20HE%20in%20the%20revising%20stage.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=917>) · Original paper and literature note are not hosted.
### P205 - Fig. 4
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** Figure 4 Comparing process maps of the revising stage between CL and CN groups
**Purpose:** Distinct appendix checklist/control process comparison, including complete lower caption. **Visual forms:** Process network.
**Physical source pages:** 38.
[PNG](<07_Mechanisms_and_Heterogeneity/P205-Fig04-Comparing%20process%20maps%20of%20the%20revising%20stage%20between%20CL%20and%20CN%20groups.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=918>) · Original paper and literature note are not hosted.
### P205 - Fig. 5
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** Figure 5 Comparing process maps of the revising stage between HE and CL groups
**Purpose:** Distinct appendix expert/checklist process comparison. **Visual forms:** Process network.
**Physical source pages:** 38.
[PNG](<07_Mechanisms_and_Heterogeneity/P205-Fig05-Comparing%20process%20maps%20of%20the%20revising%20stage%20between%20HE%20and%20CL%20groups.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=919>) · Original paper and literature note are not hosted.
### P205 - Fig. 6
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** Figure 6 Comparing process maps of the revising stage between HE and CN groups
**Purpose:** Distinct appendix expert/control process comparison. **Visual forms:** Process network.
**Physical source pages:** 39.
[PNG](<07_Mechanisms_and_Heterogeneity/P205-Fig06-Comparing%20process%20maps%20of%20the%20revising%20stage%20between%20HE%20and%20CN%20groups.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=920>) · Original paper and literature note are not hosted.
### P206 - Table 6
**Paper:** Beyond traditional interviews: Psychometric analysis of asynchronous video interviews for personality and interview performance evaluation using machine learning
**Original caption:** Table 6 The proportion of variance explained (R²) by the combined audio, visual, and verbal modalities for each of the interview questions, as well as for the aggregated Extraversion and Conscientiousness questions (i.e., test of Trait Activation Theory hypotheses). Notes. Values are R2; X = Extraversion; C = Conscientiousness; SSE = Social self-esteem; SB = Social boldness; Soc = Sociability; Liv = Liveliness; Org = Organization; Dil = Diligence; Per = Perfectionism; Pru = Prudence; the individual question that explained the largest proportion of variance per variable is boldfaced; hypothesized TAT relations are grey-shaded; negative values signify the inefficiency of the algorithm to predict the target variable; the column “All 8 questions” is included for comparison purposes, and reports the results when all eight interview questions are merged together; for R2 > 0.01, p < .05; for R2 > 0.02, p < .01; and for R2 > 0.03, p < .001.
**Purpose:** Gray-shaded cells distinguish trait-relevant questions and show trait activation heterogeneity. **Visual forms:** Complex table.
**Physical source pages:** 10.
[PNG](<07_Mechanisms_and_Heterogeneity/P206-Table06-The%20proportion%20of%20variance%20explained%20(R2)%20by%20the%20combined%20audio,%20visual,%20and%20verbal%20modalities%20for%20each%20of%20the%20interview%20questions,%20as%20well%20as%20for%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=921>) · Original paper and literature note are not hosted.
### P206 - Table 7
**Paper:** Beyond traditional interviews: Psychometric analysis of asynchronous video interviews for personality and interview performance evaluation using machine learning
**Original caption:** Table 7 The proportion of variance explained (R²) by each of the audio, visual, and verbal modalities for the aggregated Extraversion and Conscientiousness questions (i.e., test of Trait Activation Theory hypotheses). Notes. Values are R²; the questions that explained the largest proportion of variance per variable are boldfaced; hypothesized TAT relations are grey-shaded; negative values signify the inefficiency of the algorithm to predict the target variable; the column “All 8 questions” is included for comparison purposes, and reports the results when all eight interview questions are merged together; for R² > 0.01, p < .05; for R² > 0.02, p < .01; and for R² > 0.03, p < .001.
**Purpose:** Visually compares activated traits across modalities and self/observer assessments. **Visual forms:** Complex table.
**Physical source pages:** 11.
[PNG](<07_Mechanisms_and_Heterogeneity/P206-Table07-The%20proportion%20of%20variance%20explained%20(R2)%20by%20each%20of%20the%20audio,%20visual,%20and%20verbal%20modalities%20for%20the%20aggregated%20Extraversion%20and%20Conscientiousness%20questions....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=922>) · Original paper and literature note are not hosted.
### P210 - Fig. 2
**Paper:** Cognitive Offloading
**Original caption:** Figure 2 Caption. Relation between Internal Demand and Cognitive Offloading. There exists a consistent relation between the amount of internal demand, as indexed in a condition where offloading is restricted, and the amount of spontaneous offloading behavior observed in a condition where the behavior is not restricted. This has been demonstrated across a number of different domains. With respect to external normalization (Panel A) as the internal costs of stimulus rotation increase when individuals are forced to remain upright (i.e., see rotation costs in ms/degree in “Restricted Motion Conditions”; larger values represent greater costs), the likelihood that an individual spontaneously physical rotates increases (i.e., see “Free Rotation Conditions”; larger values represent a higher frequency of offloading; [9]). In intention offloading (Panel B) and short-term memory (Panel C) as the unaided memory performance decreases (see “Accuracy without intention offloading” in Panel B and “Accuracy: No Choice” in Panel C), the likelihood that an individual spontaneously offloads the memory demands into the environment (i.e., setting reminders; writing the to-be-remembered items down) increases (see “Intentions offloaded when possible” in Panel B and “Choice Behavior” in Panel C; in both cases higher values represent a higher frequency of offloading; [10, 29]). Graphs modified with permission from [9], [29], and [45].
**Purpose:** Evidence for demand-dependent strategy choice across three experimental domains; all original panels and detached full caption retained. **Visual forms:** Composite, Line plot, Bar chart.
**Physical source pages:** 28, 36.
[PNG](<07_Mechanisms_and_Heterogeneity/P210-Fig02-Caption.%20Relation%20between%20Internal%20Demand%20and%20Cognitive%20Offloading.%20There%20exists%20a%20consistent%20relation%20between%20the%20amount%20of%20internal%20demand,%20as%20indexed%20in%20a....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=923>) · Original paper and literature note are not hosted.
### P213 - Fig. 2
**Paper:** COVID-19 school closures, learning losses and intergenerational mobility
**Original caption:** Fig. 2 Composition of learning modalities during school closures by country income group. Authors’ estimates based on HFPS data from a sample of 30 developing countries, including 5 LICs, 11 LMICs, 12 UMICs and 2 HICs.
**Purpose:** Differences in access to learning modalities across country income groups. **Visual forms:** Stacked bar chart.
**Physical source pages:** 5.
[PNG](<07_Mechanisms_and_Heterogeneity/P213-Fig02-Composition%20of%20learning%20modalities%20during%20school%20closures%20by%20country%20income%20group.%20Authors%E2%80%99%20estimates%20based%20on%20HFPS%20data%20from%20a%20sample%20of%2030%20developing%20countries,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=924>) · Original paper and literature note are not hosted.
### P213 - Fig. 3
**Paper:** COVID-19 school closures, learning losses and intergenerational mobility
**Original caption:** Fig. 3 Distribution of learning modalities during school closures, by level of parental education. Authors’ estimates based on HFPS data from a sample of 30 developing countries, including 5 LICs, 11 LMICs, 12 UMICs and 2 HICs.
**Purpose:** Recovered by direct visual comparison with source page. **Visual forms:** Stacked bar chart.
**Physical source pages:** 5.
[PNG](<07_Mechanisms_and_Heterogeneity/P213-Fig03-Distribution%20of%20learning%20modalities%20during%20school%20closures,%20by%20level%20of%20parental%20education.%20Authors%E2%80%99%20estimates%20based%20on%20HFPS%20data%20from%20a%20sample%20of%2030%20developing....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=925>) · Original paper and literature note are not hosted.
### P213 - Fig. 6
**Paper:** COVID-19 school closures, learning losses and intergenerational mobility
**Original caption:** Fig. 6 Share of children with improved education over their parents under different scenarios. Authors’ estimates based on data from Azevedo et al. (2022) and GDIM. Distributional losses based on Scenario 1 assumptions on the loss index. These results are not affected by alternative loss index assumptions for the vast majority of countries, with the exception of Guatemala and Chile, where absolute mobility declines further in alternative scenarios. We thus omit alternative scenarios from the graph for ease of legibility.
**Purpose:** Country-specific heterogeneity in absolute mobility losses, including distributional losses. **Visual forms:** Dot plot.
**Physical source pages:** 8.
[PNG](<07_Mechanisms_and_Heterogeneity/P213-Fig06-Share%20of%20children%20with%20improved%20education%20over%20their%20parents%20under%20different%20scenarios.%20Authors%E2%80%99%20estimates%20based%20on%20data%20from%20Azevedo%20et%20al.%20(2022)%20and%20GDIM....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=926>) · Original paper and literature note are not hosted.
### P213 - App Fig. A1.2
**Paper:** COVID-19 school closures, learning losses and intergenerational mobility
**Original caption:** Figure A1.2: Learning Losses simulations by different scenarios by LAYS components and main channels. (a) By income level. (b) By region (excluding High-income countries). Source: Authors’ calculations.
**Purpose:** Simulation decomposition compares quantity, quality and dropout channels across income groups and regions. **Visual forms:** Stacked bar chart, Composite.
**Physical source pages:** 3.
[PNG](<07_Mechanisms_and_Heterogeneity/P213-App-FigA1.2-Learning%20Losses%20simulations%20by%20different%20scenarios%20by%20LAYS%20components%20and%20main%20channels.%20(a)%20By%20income%20level.%20(b)%20By%20region%20(excluding%20High-income....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=927>) · Original paper and literature note are not hosted.
### P214 - Fig. 2
**Paper:** Developing measures of engagement in PISA
**Original caption:** Figure 2. Differences in disengagement by gender (girls – boys). Notes: Values represent differences between OECD standardised measures (mean and standard deviation of 0 and 1, respectively, across OECD countries). Negative differences (girls-boys) indicate higher disengagement for boys. Rapid guessing is only available for countries/economies that administered computer-based assessment.
**Purpose:** Country-level gender differences in disengagement. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 26.
[PNG](<07_Mechanisms_and_Heterogeneity/P214-Fig02-Differences%20in%20disengagement%20by%20gender%20(girls%20-%20boys).%20Notes%20Values%20represent%20differences%20between%20OECD%20standardised%20measures%20(mean%20and%20standard%20deviation%20of%200%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=928>) · Original paper and literature note are not hosted.
### P214 - Fig. 3
**Paper:** Developing measures of engagement in PISA
**Original caption:** Figure 3. Differences in disengagement by socio-economic status (top quarter - bottom quarter). Notes: Values represent differences between OECD standardised measures (mean and standard deviation of 0 and 1, respectively, across OECD countries). Negative differences (top quarter - bottom quarter) indicate higher disengagement for the bottom quarter on the ESCS. Rapid guessing is only available for countries/economies that administered computer-based assessment.
**Purpose:** Country-level socio-economic subgroup differences in disengagement. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 27.
[PNG](<07_Mechanisms_and_Heterogeneity/P214-Fig03-Differences%20in%20disengagement%20by%20socio-economic%20status%20(top%20quarter%20-%20bottom%20quarter).%20Notes%20Values%20represent%20differences%20between%20OECD%20standardised%20measures%20(mean....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=929>) · Original paper and literature note are not hosted.
### P214 - Fig. 4
**Paper:** Developing measures of engagement in PISA
**Original caption:** Figure 4. Differences in disengagement by immigration background (non-immigrant - immigrant background). Notes: Values represent differences between OECD standardised measures (mean and standard deviation of 0 and 1, respectively, across OECD countries). Negative differences (non-immigrant - immigrant background) indicate higher disengagement for students with an immigrant background. Rapid guessing is only available for countries/economies that administered computer-based assessment.
**Purpose:** Country-level differences in disengagement by immigration background. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 28.
[PNG](<07_Mechanisms_and_Heterogeneity/P214-Fig04-Differences%20in%20disengagement%20by%20immigration%20background%20(non-immigrant%20-%20immigrant%20background).%20Notes%20Values%20represent%20differences%20between%20OECD%20standardised....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=930>) · Original paper and literature note are not hosted.
### P216 - Fig. 1
**Paper:** Does geographical location impact educational disparities among Ecuadorians? A novel two-stage inequality decomposition method
**Original caption:** Fig. 1 ‘Years of schooling’ distribution by province (2014 and 2021). The lines in each graph correspond to the kernel density function of the variable in each province for the study years. The Gaussian kernel function has been used (Silverman, 1986). SOURCE: Authors’ calculations based on microdata bases of Ecuador’s National Survey of Employment, Unemployment and Underemployment for 2014 and 2021 (Instituto Nacional de Estadística y Censos [INEC], 2022a).
**Purpose:** Provincial and temporal heterogeneity in schooling distributions or inequality contributions. **Visual forms:** Distribution, Composite.
**Physical source pages:** 9.
[PNG](<07_Mechanisms_and_Heterogeneity/P216-Fig01-%E2%80%98Years%20of%20schooling%E2%80%99%20distribution%20by%20province%20(2014%20and%202021).%20The%20lines%20in%20each%20graph%20correspond%20to%20the%20kernel%20density%20function%20of%20the%20variable%20in%20each%20province....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=931>) · Original paper and literature note are not hosted.
### P216 - Fig. 2
**Paper:** Does geographical location impact educational disparities among Ecuadorians? A novel two-stage inequality decomposition method
**Original caption:** Fig. 2 Provincial contribution to the within provinces component of Ecuadorians educational inequality (2014 and 2021). The provinces are grouped by quartiles according to their contribution. SOURCE: Authors’ calculation based on the microdata bases of Ecuador’s National Survey of Employment, Unemployment and Underemployment for 2014 and 2021 (Instituto Nacional de Estadística y Censos [INEC], 2022a). The geographic representation was conducted using a GIS (Geographic Information Systems) mapping programme for which the corresponding license for use is available. A shapefile with the political limits of the provinces of Ecuador has been used in the preparation of the map. This shapefile is freely available from the website of the University of Azuay (https://gis.uazuay.edu.ec/descargas/promsa/ecuador.htm).
**Purpose:** Provincial and temporal heterogeneity in schooling distributions or inequality contributions. **Visual forms:** Map, Composite.
**Physical source pages:** 11.
[PNG](<07_Mechanisms_and_Heterogeneity/P216-Fig02-Provincial%20contribution%20to%20the%20within%20provinces%20component%20of%20Ecuadorians%20educational%20inequality%20(2014%20and%202021).%20The%20provinces%20are%20grouped%20by%20quartiles%20according....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=932>) · Original paper and literature note are not hosted.
### P217 - Fig. 2
**Paper:** Do smartphone usage scales predict behavior?
**Original caption:** Figure 2. Means of high (N=92) (cluster 1) and low users (N=146) (cluster 2) derived from objective data following a k-means cluster analysis.
**Purpose:** Describes high- versus low-use behavioral clusters used for classification validation. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 16, 17.
[PNG](<07_Mechanisms_and_Heterogeneity/P217-Fig02-Means%20of%20high%20(N=92)%20(cluster%201)%20and%20low%20users%20(N=146)%20(cluster%202)%20derived%20from%20objective%20data%20following%20a%20k-means%20cluster%20analysis.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=933>) · Original paper and literature note are not hosted.
### P221 - Fig. 1
**Paper:** From asking to observing. Behavioural measures of socio-emotional and motivational skills in large-scale assessments
**Original caption:** Fig. 1. Gender differences in PSEC indicators. Notes: The gender gap refers to the difference between females and males. A negative number implies that females have lower mean values than males. A positive value implies that females have higher mean values than males. An asterisk (*) next to an indicator name indicates that the estimated d value is statistically significant at least at the 5% level. Data for all countries in the sample can be consulted in Table H2 in the Online Supplementary Annex. Source: PISA 2012data.
**Purpose:** Compares demographic group differences across candidate PSEC measures. **Visual forms:** Bar chart.
**Physical source pages:** 12.
[PNG](<07_Mechanisms_and_Heterogeneity/P221-Fig01-Gender%20differences%20in%20PSEC%20indicators.%20Notes%20The%20gender%20gap%20refers%20to%20the%20difference%20between%20females%20and%20males.%20A%20negative%20number%20implies%20that%20females%20have%20lower....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=934>) · Original paper and literature note are not hosted.
### P221 - Fig. 2
**Paper:** From asking to observing. Behavioural measures of socio-emotional and motivational skills in large-scale assessments
**Original caption:** Fig. 2. Socio-economic differences in PSEC indicators. Notes: The socio-economic (SES) gap refers to the difference between socio-economically advantaged and socio-economically disadvantaged students. PISA contains an aggregate SES measure based on students’ reports on their parents’ educational attainment, occupational status and availability of economic and cultural resources in their home. Advantaged students are students in the top 25% of the country specific distribution of SES. Disadvantages students are students in the bottom 25% of the country specific distribution of SES. A negative number implies that advantaged have lower mean values than disadvantaged students. A positive value implies that advantaged have higher mean values than disadvantaged students. An asterisk (*) next to an indicator name indicates that the estimated d value is statistically significant at least at the 5% level. Data for all countries in the sample can be consulted in Table I2 in the Online Supplementary Annex. Source: PISA 2012data.
**Purpose:** Compares demographic group differences across candidate PSEC measures. **Visual forms:** Bar chart.
**Physical source pages:** 13.
[PNG](<07_Mechanisms_and_Heterogeneity/P221-Fig02-Socio-economic%20differences%20in%20PSEC%20indicators.%20Notes%20The%20socio-economic%20(SES)%20gap%20refers%20to%20the%20difference%20between%20socio-economically%20advantaged%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=935>) · Original paper and literature note are not hosted.
### P221 - Fig. 3
**Paper:** From asking to observing. Behavioural measures of socio-emotional and motivational skills in large-scale assessments
**Original caption:** Fig. 3. Differences by immigrant background in PSEC indicators. Notes: The immigrant gap refers to the difference between students with and students without an immigrant background. A negative number implies that immigrant students have lower mean values than students without an immigrant background. A positive value implies that immigrant students have higher mean values than students without an immigrant background. An asterisk (*) next to an indicator name indicates that the estimated d value is statistically significant at least at the 5% level. Data for all countries in the sample can be consulted in Table I2 in the Online Supplementary Annex. Source: PISA 2012data.
**Purpose:** Compares demographic group differences across candidate PSEC measures. **Visual forms:** Bar chart.
**Physical source pages:** 14.
[PNG](<07_Mechanisms_and_Heterogeneity/P221-Fig03-Differences%20by%20immigrant%20background%20in%20PSEC%20indicators.%20Notes%20The%20immigrant%20gap%20refers%20to%20the%20difference%20between%20students%20with%20and%20students%20without%20an%20immigrant....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=936>) · Original paper and literature note are not hosted.
### P224 - Fig. 2
**Paper:** Generative AI without guardrails can harm learning: Evidence from high school mathematics
**Original caption:** Fig. 2. Fraction of times GPT Base returns an incorrect answer due to a logical (Left) or arithmetic (Right) error for 57 practice problems across grades and sessions. (Note that GPT Tutor does not return an answer by design, and is further given the correct answer in the prompt).
**Purpose:** Audits model error types as evidence for the proposed learning mechanism; retain both panels and full note. **Visual forms:** Histogram, Composite.
**Physical source pages:** 5.
[PNG](<07_Mechanisms_and_Heterogeneity/P224-Fig02-Fraction%20of%20times%20GPT%20Base%20returns%20an%20incorrect%20answer%20due%20to%20a%20logical%20(Left)%20or%20arithmetic%20(Right)%20error%20for%2057%20practice%20problems%20across%20grades%20and%20sessions....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=937>) · Original paper and literature note are not hosted.
### P224 - Fig. 3
**Paper:** Generative AI without guardrails can harm learning: Evidence from high school mathematics
**Original caption:** Fig. 3. Student engagement—given by (A) average number of student messages per problem, and (B) average fraction of student session conversations that have no superficial messages (simply restating the question or asking for the answer) per session—by treatment (GPT Base and GPT Tutor) over time.
**Purpose:** Engagement differences investigate reliance on answer copying; both panels and caption retained. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P224-Fig03-Student%20engagement-given%20by%20(A)%20average%20number%20of%20student%20messages%20per%20problem,%20and%20(B)%20average%20fraction%20of%20student%20session%20conversations%20that%20have%20no%20superficial....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=938>) · Original paper and literature note are not hosted.
### P227 - Fig. 3
**Paper:** How do test-takers rate their effort? A comparative analysis of self-report and log file data
**Original caption:** Fig. 3. Student profiles based on time spent on task, number of clicks, score achieved and self-reported effort (TOT: total time on task; CLICK: total number of clicks; SCORE: test score; SRE: self-reported effort).
**Purpose:** Compares effort/performance profiles across the three clusters. **Visual forms:** Bar chart.
**Physical source pages:** 30.
[PNG](<07_Mechanisms_and_Heterogeneity/P227-Fig03-Student%20profiles%20based%20on%20time%20spent%20on%20task,%20number%20of%20clicks,%20score%20achieved%20and%20self-reported%20effort%20(TOT%20total%20time%20on%20task;%20CLICK%20total%20number%20of%20clicks;....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=939>) · Original paper and literature note are not hosted.
### P230 - Fig. 5
**Paper:** Impact of AI assistance on student agency
**Original caption:** Fig. 5. Comparing the control group (AI) with the three experimental groups (NR, SR, and SAI) using six measures for each course.
**Purpose:** Course-specific treatment comparisons assess consistency and heterogeneity; all ten course composites remain whole. **Visual forms:** Violin plot, Boxplot, Composite.
**Physical source pages:** 15.
[PNG](<07_Mechanisms_and_Heterogeneity/P230-Fig05-Comparing%20the%20control%20group%20(AI)%20with%20the%20three%20experimental%20groups%20(NR,%20SR,%20and%20SAI)%20using%20six%20measures%20for%20each%20course.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=940>) · Original paper and literature note are not hosted.
### P231 - Fig. 2
**Paper:** Inequality of opportunity in non-cognitive skills: a cross-region analysis
**Original caption:** Fig. 2 CDF of dependent variable based on parental education and family SES (Suzhou).
**Purpose:** Distributional differences in conscientiousness across parental-education and family-SES groups in the two illustrative sites. **Visual forms:** Cumulative distribution, Composite.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P231-Fig02-CDF%20of%20dependent%20variable%20based%20on%20parental%20education%20and%20family%20SES%20(Suzhou).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=941>) · Original paper and literature note are not hosted.
### P231 - Fig. 3
**Paper:** Inequality of opportunity in non-cognitive skills: a cross-region analysis
**Original caption:** Fig. 3 CDF of dependent variable based on parental education and family SES (Houston).
**Purpose:** Distributional differences in conscientiousness across parental-education and family-SES groups in the two illustrative sites. **Visual forms:** Cumulative distribution, Composite.
**Physical source pages:** 8.
[PNG](<07_Mechanisms_and_Heterogeneity/P231-Fig03-CDF%20of%20dependent%20variable%20based%20on%20parental%20education%20and%20family%20SES%20(Houston).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=942>) · Original paper and literature note are not hosted.
### P233 - Fig. 1
**Paper:** Is grit persistence adaptive? Goal pursuit behavior when faced with a difficult goal
**Original caption:** Figure 1 Effect of grit and importance of the goal (high versus low) on goal disengagement. CI, confidence interval; SD: standard deviation
**Purpose:** Simple-slope plot of goal-importance or time-limit moderation of grit associations. **Visual forms:** Line/profile plot.
**Physical source pages:** 14.
[PNG](<07_Mechanisms_and_Heterogeneity/P233-Fig01-Effect%20of%20grit%20and%20importance%20of%20the%20goal%20(high%20versus%20low)%20on%20goal%20disengagement.%20CI,%20confidence%20interval;%20SD%20standard%20deviation.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=943>) · Original paper and literature note are not hosted.
### P233 - Fig. 2
**Paper:** Is grit persistence adaptive? Goal pursuit behavior when faced with a difficult goal
**Original caption:** Figure 2 Effect of grit and importance of the goal (high versus low) on goal content adjustment. CI, confidence interval; SD: standard deviation
**Purpose:** Simple-slope plot of goal-importance or time-limit moderation of grit associations. **Visual forms:** Line/profile plot.
**Physical source pages:** 15.
[PNG](<07_Mechanisms_and_Heterogeneity/P233-Fig02-Effect%20of%20grit%20and%20importance%20of%20the%20goal%20(high%20versus%20low)%20on%20goal%20content%20adjustment.%20CI,%20confidence%20interval;%20SD%20standard%20deviation.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=944>) · Original paper and literature note are not hosted.
### P233 - Fig. 3
**Paper:** Is grit persistence adaptive? Goal pursuit behavior when faced with a difficult goal
**Original caption:** Figure 3 Effect of grit and importance of the goal (high versus low) on goal level adjustment. CI, confidence interval; SD: standard deviation
**Purpose:** Simple-slope plot of goal-importance or time-limit moderation of grit associations. **Visual forms:** Line/profile plot.
**Physical source pages:** 16.
[PNG](<07_Mechanisms_and_Heterogeneity/P233-Fig03-Effect%20of%20grit%20and%20importance%20of%20the%20goal%20(high%20versus%20low)%20on%20goal%20level%20adjustment.%20CI,%20confidence%20interval;%20SD%20standard%20deviation.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=945>) · Original paper and literature note are not hosted.
### P233 - Fig. 4
**Paper:** Is grit persistence adaptive? Goal pursuit behavior when faced with a difficult goal
**Original caption:** Figure 4 Effect of grit and time limit on time spent working on unsolvable tasks; CI, confidence interval
**Purpose:** Simple-slope plot of goal-importance or time-limit moderation of grit associations. **Visual forms:** Line/profile plot.
**Physical source pages:** 21.
[PNG](<07_Mechanisms_and_Heterogeneity/P233-Fig04-Effect%20of%20grit%20and%20time%20limit%20on%20time%20spent%20working%20on%20unsolvable%20tasks;%20CI,%20confidence%20interval.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=946>) · Original paper and literature note are not hosted.
### P235 - Fig. 4
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** Figure 4: Proportion of questions answered by question and treatment group
**Purpose:** Item-position pattern of the incentive effect on attempted responses. **Visual forms:** Line/profile plot.
**Physical source pages:** 30.
[PNG](<07_Mechanisms_and_Heterogeneity/P235-Fig04-Proportion%20of%20questions%20answered%20by%20question%20and%20treatment%20group.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=947>) · Original paper and literature note are not hosted.
### P235 - Fig. 5
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** Figure 5: Treatment effects by predicted score, U.S. Notes: We predict score using age, gender, race/ethnicity and baseline exam score in the U.S. control group. We estimate the control and treatment lines using kernel weighting. The vertical line at 14.15 is the U.S. national average.
**Purpose:** Heterogeneity in experimental performance and effort effects by predicted prior performance. **Visual forms:** Scatter, Composite.
**Physical source pages:** 31.
[PNG](<07_Mechanisms_and_Heterogeneity/P235-Fig05-Treatment%20effects%20by%20predicted%20score,%20U.S.%20Notes%20We%20predict%20score%20using%20age,%20gender,%20race%20ethnicity%20and%20baseline%20exam%20score%20in%20the%20U.S.%20control%20group.%20We%20estimate....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=948>) · Original paper and literature note are not hosted.
### P235 - Fig. A1
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** Figure A1: Treatment effects by predicted score, Shanghai. Notes: We predict score using age, gender, and baseline exam score in the Shanghai control group. We estimate the control and treatment lines using kernel weighting.
**Purpose:** Shanghai counterpart of the predicted-score heterogeneity analysis. **Visual forms:** Scatter, Composite.
**Physical source pages:** 38.
[PNG](<07_Mechanisms_and_Heterogeneity/P235-FigA1-Treatment%20effects%20by%20predicted%20score,%20Shanghai.%20Notes%20We%20predict%20score%20using%20age,%20gender,%20and%20baseline%20exam%20score%20in%20the%20Shanghai%20control%20group.%20We%20estimate%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=949>) · Original paper and literature note are not hosted.
### P242 - Fig. 4
**Paper:** Self-Regulated Learning: Beliefs, Techniques, and Illusions
**Original caption:** Figure 4. Percentage recalled (a) and mean judgment of learning (b) as a function of the latency of cued recall. Latency (response time in quartiles) is presented from fastest (1) to slowest (4). From Benjamin et al. (1998).
**Purpose:** Evidence that retrieval fluency affects confidence differently from subsequent recall. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 15.
[PNG](<07_Mechanisms_and_Heterogeneity/P242-Fig04-Percentage%20recalled%20(a)%20and%20mean%20judgment%20of%20learning%20(b)%20as%20a%20function%20of%20the%20latency%20of%20cued%20recall.%20Latency%20(response%20time%20in%20quartiles)%20is%20presented%20from....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=950>) · Original paper and literature note are not hosted.
### P242 - Fig. 5
**Paper:** Self-Regulated Learning: Beliefs, Techniques, and Illusions
**Original caption:** Figure 5. Number of participants (out of 120) who judged massing as more effective than, equally effective as, or less effective than interleaving in Kornell & Bjork (2008a). For each judgment, participants are divided into groups on the basis of their actual performance.
**Purpose:** Contrasts judged and actual effectiveness of practice schedules. **Visual forms:** Stacked bar chart.
**Physical source pages:** 18.
[PNG](<07_Mechanisms_and_Heterogeneity/P242-Fig05-Number%20of%20participants%20(out%20of%20120)%20who%20judged%20massing%20as%20more%20effective%20than,%20equally%20effective%20as,%20or%20less%20effective%20than%20interleaving%20in%20Kornell%20&%20Bjork....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=951>) · Original paper and literature note are not hosted.
### P244 - Fig. 1
**Paper:** Speeding in Web Surveys: The tendency to answer very fast and its association with straightlining
**Original caption:** Figure 1. Percentage of persistent speeders by age groups
**Purpose:** Shows age heterogeneity in persistent speeding. **Visual forms:** Bar chart.
**Physical source pages:** 4.
[PNG](<07_Mechanisms_and_Heterogeneity/P244-Fig01-Percentage%20of%20persistent%20speeders%20by%20age%20groups.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=952>) · Original paper and literature note are not hosted.
### P244 - Fig. 2
**Paper:** Speeding in Web Surveys: The tendency to answer very fast and its association with straightlining
**Original caption:** Figure 2. Number of grid questions with straightlining answers by speeding tendency and education
**Purpose:** Shows education moderation of the association between speeding and straightlining. **Visual forms:** Forest/coefficient plot.
**Physical source pages:** 5.
[PNG](<07_Mechanisms_and_Heterogeneity/P244-Fig02-Number%20of%20grid%20questions%20with%20straightlining%20answers%20by%20speeding%20tendency%20and%20education.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=953>) · Original paper and literature note are not hosted.
### P245 - Fig. 2
**Paper:** Students’ subjective well-being, school bullying, and belonging during the COVID-19 pandemic: Comparison between PISA 2018 and PISA 2022
**Original caption:** Fig. 2 Analysis of the relationship between mediating variables in the PISA 2018 data. a represents meditating analysis for PISA 2018; (b) represents meditating analysis for PISA 2022.
**Purpose:** Estimated school-belonging mediation pathways linking bullying and subjective well-being, presented separately for both survey cycles. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 7.
[PNG](<07_Mechanisms_and_Heterogeneity/P245-Fig02-Analysis%20of%20the%20relationship%20between%20mediating%20variables%20in%20the%20PISA%202018%20data.%20a%20represents%20meditating%20analysis%20for%20PISA%202018;%20(b)%20represents%20meditating%20analysis....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=954>) · Original paper and literature note are not hosted.
### P247 - Fig. 1
**Paper:** Taking PISA Seriously: How Accurate are Low-Stakes Exams?
**Original caption:** Fig. 1 Standardized Time for Serious and Non-serious Students Note: Data Source: 2015 PISA Cognitive item dataset. Time spent on each question (by all students who are faced with the question and who spend some time on it, whether or not they answer it) is standardized so it has mean zero and variance 1. For each position in a cluster, the median standardized time of the questions in that position is calculated. The y-axis depicts the median time spent on items in each order
**Purpose:** Compares response behavior across disengagement types or decomposes country gains into prevalence, intensity and ability components. **Visual forms:** Line/profile plot, Scatter.
**Physical source pages:** 15.
[PNG](<07_Mechanisms_and_Heterogeneity/P247-Fig01-Standardized%20Time%20for%20Serious%20and%20Non-serious%20Students%20Note%20Data%20Source%202015%20PISA%20Cognitive%20item%20dataset.%20Time%20spent%20on%20each%20question%20(by%20all%20students%20who%20are....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=955>) · Original paper and literature note are not hosted.
### P247 - Fig. 2
**Paper:** Taking PISA Seriously: How Accurate are Low-Stakes Exams?
**Original caption:** Fig. 2 Standardized Score for Serious and Non-serious Students Note: Data Source: 2015 PISA Cognitive item dataset. The score for each question, 0, 0.5 or 1, is standardized so the overall score has mean zero and variance 1. Items that are not reached or missing are dropped from the sample. The no response items are assigned a score of 0. For each position in a cluster, the average standardized score of all questions in that position is calculated. The y-axis depicts the mean standardized score of the items in each order
**Purpose:** Compares response behavior across disengagement types or decomposes country gains into prevalence, intensity and ability components. **Visual forms:** Line/profile plot, Scatter.
**Physical source pages:** 15.
[PNG](<07_Mechanisms_and_Heterogeneity/P247-Fig02-Standardized%20Score%20for%20Serious%20and%20Non-serious%20Students%20Note%20Data%20Source%202015%20PISA%20Cognitive%20item%20dataset.%20The%20score%20for%20each%20question,%200,%200.5%20or%201,%20is....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=956>) · Original paper and literature note are not hosted.
### P247 - Fig. 3
**Paper:** Taking PISA Seriously: How Accurate are Low-Stakes Exams?
**Original caption:** Fig. 3 Time for Correct and Incorrect Answers for Serious and Non-serious Students Note: Data Source: 2015 PISA Cognitive item dataset. The residuals of time spent for each student and question are obtained by running a regression of time spent on each item on type of question (multiple choice or open-ended), position within a cluster and position of the cluster and getting the residuals. Here time spent is conditional on having answered the question. The y-axis depicts the mean of the residual time relative to the difficulty of the items which is measured by the fraction who got the question correct. The green line is for non-serious students and the red line is for serious students
**Purpose:** Compares response behavior across disengagement types or decomposes country gains into prevalence, intensity and ability components. **Visual forms:** Line/profile plot, Scatter.
**Physical source pages:** 17.
[PNG](<07_Mechanisms_and_Heterogeneity/P247-Fig03-Time%20for%20Correct%20and%20Incorrect%20Answers%20for%20Serious%20and%20Non-serious%20Students%20Note%20Data%20Source%202015%20PISA%20Cognitive%20item%20dataset.%20The%20residuals%20of%20time%20spent%20for....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=957>) · Original paper and literature note are not hosted.
### P247 - Fig. 4
**Paper:** Taking PISA Seriously: How Accurate are Low-Stakes Exams?
**Original caption:** Fig. 4 Time for Correct and Incorrect Answers for Serious and Missing-item Students Note: Data Source: 2015 PISA Cognitive item dataset. The residuals of time spent for each student and question are obtained by running a regression of time spent on each item on type of question (multiple choice or open-ended), position within a cluster and position of the cluster and getting the residuals. Here time spent is conditional on having answered the question. The y-axis depicts the mean of the residual time relative to the difficulty of the items which is measured by the fraction who got the question correct. The red line is for serious students while the black line is for non-serious students who satisfy criterion 3, or missing-item students
**Purpose:** Compares response behavior across disengagement types or decomposes country gains into prevalence, intensity and ability components. **Visual forms:** Line/profile plot, Scatter.
**Physical source pages:** 18.
[PNG](<07_Mechanisms_and_Heterogeneity/P247-Fig04-Time%20for%20Correct%20and%20Incorrect%20Answers%20for%20Serious%20and%20Missing-item%20Students%20Note%20Data%20Source%202015%20PISA%20Cognitive%20item%20dataset.%20The%20residuals%20of%20time%20spent%20for....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=958>) · Original paper and literature note are not hosted.
### P247 - Fig. 5
**Paper:** Taking PISA Seriously: How Accurate are Low-Stakes Exams?
**Original caption:** Fig. 5 Time for Correct and Incorrect Answers for Serious and Non-serious Students After Removing Missing-item Students Note: Data Source: 2015 PISA Cognitive item dataset. The residuals of time spent for each student and question are obtained by running a regression of time spent on each item on type of question (multiple choice or open-ended), position within a cluster and position of the cluster and getting the residuals. Here time spent is conditional on having answered the question. The y-axis depicts the mean of the residual time relative to the difficulty of the items which is measured by the fraction who got the question correct. The red line is for serious students while the black line is for non-serious students excluding missing-item students
**Purpose:** Compares response behavior across disengagement types or decomposes country gains into prevalence, intensity and ability components. **Visual forms:** Line/profile plot, Scatter.
**Physical source pages:** 19.
[PNG](<07_Mechanisms_and_Heterogeneity/P247-Fig05-Time%20for%20Correct%20and%20Incorrect%20Answers%20for%20Serious%20and%20Non-serious%20Students%20After%20Removing%20Missing-item%20Students%20Note%20Data%20Source%202015%20PISA%20Cognitive%20item....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=959>) · Original paper and literature note are not hosted.
### P247 - Fig. 6
**Paper:** Taking PISA Seriously: How Accurate are Low-Stakes Exams?
**Original caption:** Fig. 6 yns Versus its Components for Non-Serious Students Note:The top panel plots the scatter plot and regression lines between ln(ans) and ln(yns), showing the contribution of non-serious students’ ability to the increased fraction correct. The middle panel plots the relationship between ln(ens) and ln(yns) for every country, showing the contribution of extent of non-seriousness. The bottom panel plots the relation-ship between ln(pns) and ln(yns) for every country, showing the contribution of proportion of no-serious students
**Purpose:** Compares response behavior across disengagement types or decomposes country gains into prevalence, intensity and ability components. **Visual forms:** Scatter, Composite.
**Physical source pages:** 37.
[PNG](<07_Mechanisms_and_Heterogeneity/P247-Fig06-yns%20Versus%20its%20Components%20for%20Non-Serious%20Students%20Note%20The%20top%20panel%20plots%20the%20scatter%20plot%20and%20regression%20lines%20between%20ln(ans)%20and%20ln(yns),%20showing%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=960>) · Original paper and literature note are not hosted.
### P247 - Fig. 7
**Paper:** Taking PISA Seriously: How Accurate are Low-Stakes Exams?
**Original caption:** Fig. 7 Pr(skip) and Pr(spend too little time) w.r.t. cluster and difficulty. In the figure, lowess-smoothed lines are presented. Predicted probabilities are obtained from a linear probability model with individual fixed effects as well as question characteristics such as cluster, sequence, difficulty and the type of the question. For each cluster, the predicted probability at each level of question difficulty takes the mean value of the predicted probability at that level of difficulty. In the figure, lowess-smoothed lines are presented
**Purpose:** Compares response behavior across disengagement types or decomposes country gains into prevalence, intensity and ability components. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 46.
[PNG](<07_Mechanisms_and_Heterogeneity/P247-Fig07-Pr(skip)%20and%20Pr(spend%20too%20little%20time)%20w.r.t.%20cluster%20and%20difficulty.%20In%20the%20figure,%20lowess-smoothed%20lines%20are%20presented.%20Predicted%20probabilities%20are%20obtained....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=961>) · Original paper and literature note are not hosted.
### P247 - Fig. 8
**Paper:** Taking PISA Seriously: How Accurate are Low-Stakes Exams?
**Original caption:** Fig. 8 Pr(skip) and Pr(spend too little time) w.r.t. sequence and the type of the question. Predicted prob-abilities are obtained from a linear probability model with individual fixed effects as well as question characteristics such as cluster, sequence, difficulty, and the type of the question. For each question type the mean value of the predicted probability at each order is presented
**Purpose:** Compares response behavior across disengagement types or decomposes country gains into prevalence, intensity and ability components. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 47.
[PNG](<07_Mechanisms_and_Heterogeneity/P247-Fig08-Pr(skip)%20and%20Pr(spend%20too%20little%20time)%20w.r.t.%20sequence%20and%20the%20type%20of%20the%20question.%20Predicted%20prob-abilities%20are%20obtained%20from%20a%20linear%20probability%20model%20with....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=962>) · Original paper and literature note are not hosted.
### P247 - Fig. 9
**Paper:** Taking PISA Seriously: How Accurate are Low-Stakes Exams?
**Original caption:** Fig. 9 yps Versus its Components for Partially Serious Students
**Purpose:** Compares response behavior across disengagement types or decomposes country gains into prevalence, intensity and ability components. **Visual forms:** Scatter, Composite.
**Physical source pages:** 58.
[PNG](<07_Mechanisms_and_Heterogeneity/P247-Fig09-yps%20Versus%20its%20Components%20for%20Partially%20Serious%20Students.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=963>) · Original paper and literature note are not hosted.
## 08 Robustness and Validation
### P066 - Fig. 2
**Paper:** The impact of financial education in schools on the financial literacy of Brazilian students
**Original caption:** Fig. 2. – Histogram of the propensity score. Source: Prepared by the author, based on PISA 2022 data.
**Purpose:** Propensity-score distribution diagnoses support for the PSM design. **Visual forms:** Histogram.
**Physical source pages:** 6.
[PNG](<08_Robustness_and_Validation/P066-Fig02-Histogram%20of%20the%20propensity%20score.%20Source%20Prepared%20by%20the%20author,%20based%20on%20PISA%202022%20data.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=624>) · Original paper and literature note are not hosted.
### P066 - Fig. 3
**Paper:** The impact of financial education in schools on the financial literacy of Brazilian students
**Original caption:** Fig. 3. – Kernel density of the propensity score by group. Source: Prepared by the author, based on PISA 2022 data.
**Purpose:** Propensity-score group overlap checks the support for matching. **Visual forms:** Density.
**Physical source pages:** 6.
[PNG](<08_Robustness_and_Validation/P066-Fig03-Kernel%20density%20of%20the%20propensity%20score%20by%20group.%20Source%20Prepared%20by%20the%20author,%20based%20on%20PISA%202022%20data.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=625>) · Original paper and literature note are not hosted.
### P068 - Fig. 1
**Paper:** Understanding students’ mathematical achievement through feature selection and machine learning: Evidence from PISA 2018
**Original caption:** Figure 1. Conditional SHAP feature importance rankings across PV1MATH-PV10MATH based on the top 20 variables selected by the Mutual Information method using the Random forest classifier (5x5 cross-validation)
**Purpose:** Consistency of feature importance across all ten plausible-value outcomes. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 37.
[PNG](<08_Robustness_and_Validation/P068-App-Fig01-Conditional%20SHAP%20feature%20importance%20rankings%20across%20PV1MATH-PV10MATH%20based%20on%20the%20top%2020%20variables%20selected%20by%20the%20Mutual%20Information%20method%20using%20the%20Random....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=626>) · Original paper and literature note are not hosted.
### P068 - Fig. 2
**Paper:** Understanding students’ mathematical achievement through feature selection and machine learning: Evidence from PISA 2018
**Original caption:** Figure 2. PDPs, ICE plots, and ALE plots for PV1MATH using the top Mutual information-selected features (excluding STRATUM, REPEAT, and SCHLOC).
**Purpose:** Additional predictor-response checks for the remaining selected variables. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 37, 38.
[PNG](<08_Robustness_and_Validation/P068-App-Fig02-PDPs,%20ICE%20plots,%20and%20ALE%20plots%20for%20PV1MATH%20using%20the%20top%20Mutual%20information-selected%20features%20(excluding%20STRATUM,%20REPEAT,%20and%20SCHLOC).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=627>) · Original paper and literature note are not hosted.
### P069 - Fig. A.1
**Paper:** What students learn depends on what we teach: Curriculum content coverage and student achievement in mathematics and science
**Original caption:** Fig. A.1. Forest plot of the cycle-specific and synthesized results of the total, indirect, and direct effects (weighted).
**Purpose:** Student-population-weighted sensitivity analysis. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 17.
[PNG](<08_Robustness_and_Validation/P069-FigA.1-Forest%20plot%20of%20the%20cycle-specific%20and%20synthesized%20results%20of%20the%20total,%20indirect,%20and%20direct%20effects%20(weighted).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=628>) · Original paper and literature note are not hosted.
### P074 - Fig. 6
**Paper:** Predicting Key Factors Influencing Students’ Creative Thinking: A Machine Learning Analysis Based on PISA 2022 Data from East Asia
**Original caption:** Fig. 6. LASSO regression cross-validation curve.
**Purpose:** Cross-validation diagnostic for selection of the LASSO tuning parameter. **Visual forms:** Line/profile plot.
**Physical source pages:** 10.
[PNG](<08_Robustness_and_Validation/P074-Fig06-LASSO%20regression%20cross-validation%20curve.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=629>) · Original paper and literature note are not hosted.
### P080 - Fig. B.1
**Paper:** Examining the relationship between online chatting and PISA reading literacy trends (2000-2018)
**Original caption:** Fig. B.1. Estimated trend changes, observed and predicted national reading literacy averages over 2000–2018.
**Purpose:** Observed-versus-predicted country series support model fit assessment. **Visual forms:** Composite, Line plot.
**Physical source pages:** 17, 18, 19, 20.
[PNG](<08_Robustness_and_Validation/P080-FigB.1-Estimated%20trend%20changes,%20observed%20and%20predicted%20national%20reading%20literacy%20averages%20over%202000-2018.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=630>) · Original paper and literature note are not hosted.
### P080 - Fig. E.1
**Paper:** Examining the relationship between online chatting and PISA reading literacy trends (2000-2018)
**Original caption:** Fig. E.1. Elaborated model version – standardized path coefficients and fit indices
**Purpose:** Alternative model adds control variables to probe specification sensitivity. **Visual forms:** Path diagram.
**Physical source pages:** 26.
[PNG](<08_Robustness_and_Validation/P080-FigE.1-Elaborated%20model%20version%20-%20standardized%20path%20coefficients%20and%20fit%20indices.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=631>) · Original paper and literature note are not hosted.
### P094 - Table 4
**Paper:** What teachers can do to enhance students’ well-being: Discussion
**Original caption:** Table 4. Limitations.
**Purpose:** Structured cross-study comparison of the special issue evidence, with multiple conceptual partitions. **Visual forms:** Complex table, Structured evidence comparison.
**Physical source pages:** 5.
[PNG](<08_Robustness_and_Validation/P094-Table04-Limitations.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=632>) · Original paper and literature note are not hosted.
### P104 - Fig. 1
**Paper:** Investigating factors affecting teachers’ instructional practices: Evidence from TALIS 2018 Turkish data
**Original caption:** Fig. 1. Histogram of regression standardized residual.
**Purpose:** Checks regression residual distribution or model assumptions. **Visual forms:** Residual diagnostic.
**Physical source pages:** 7.
[PNG](<08_Robustness_and_Validation/P104-Fig01-Histogram%20of%20regression%20standardized%20residual.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=633>) · Original paper and literature note are not hosted.
### P104 - Fig. 2
**Paper:** Investigating factors affecting teachers’ instructional practices: Evidence from TALIS 2018 Turkish data
**Original caption:** Fig. 2. Normal P-P Plot of regression standardized residual.
**Purpose:** Checks regression residual distribution or model assumptions. **Visual forms:** Residual diagnostic.
**Physical source pages:** 7.
[PNG](<08_Robustness_and_Validation/P104-Fig02-Normal%20P-P%20Plot%20of%20regression%20standardized%20residual.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=634>) · Original paper and literature note are not hosted.
### P104 - Fig. 3
**Paper:** Investigating factors affecting teachers’ instructional practices: Evidence from TALIS 2018 Turkish data
**Original caption:** Fig. 3. Scatter Plot of regression standardized residual.
**Purpose:** Checks regression residual distribution or model assumptions. **Visual forms:** Residual diagnostic.
**Physical source pages:** 8.
[PNG](<08_Robustness_and_Validation/P104-Fig03-Scatter%20Plot%20of%20regression%20standardized%20residual.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=635>) · Original paper and literature note are not hosted.
### P108 - Fig. A.1
**Paper:** Understanding the relationship between teacher collaboration and instructional clarity via teacher self-efficacy: A moderated mediation model of organizational commitment
**Original caption:** Fig. A.1. Three Competing Models
**Purpose:** Alternative directional models assess the robustness of the assumed mediation ordering. **Visual forms:** Path diagram, Composite.
**Physical source pages:** 11.
[PNG](<08_Robustness_and_Validation/P108-FigA.1-Three%20Competing%20Models.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=636>) · Original paper and literature note are not hosted.
### P116 - Fig. 3
**Paper:** School leadership and teacher stress: Unpacking the mediating role of job satisfaction and the employment status as a moderator
**Original caption:** Fig. 3. Q-Q plot.
**Purpose:** Normality diagnostic for Mahalanobis distances. **Visual forms:** Scatter.
**Physical source pages:** 5.
[PNG](<08_Robustness_and_Validation/P116-Fig03-Q-Q%20plot.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=637>) · Original paper and literature note are not hosted.
### P124 - Fig. U01
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** No original caption supplied.
**Purpose:** Additional Russian sample checks presented in Appendix B. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 15.
[PNG](<08_Robustness_and_Validation/P124-FigU01-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=638>) · Original paper and literature note are not hosted.
### P124 - Fig. U02
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** No original caption supplied.
**Purpose:** Multiple-imputation robustness analysis of pooled data. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 16.
[PNG](<08_Robustness_and_Validation/P124-FigU02-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=639>) · Original paper and literature note are not hosted.
### P124 - Fig. U03
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** No original caption supplied.
**Purpose:** Country-specific multiple-imputation robustness results. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 17.
[PNG](<08_Robustness_and_Validation/P124-FigU03-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=640>) · Original paper and literature note are not hosted.
### P124 - Fig. U04
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** No original caption supplied.
**Purpose:** Country-specific multiple-imputation robustness results. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 18.
[PNG](<08_Robustness_and_Validation/P124-FigU04-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=641>) · Original paper and literature note are not hosted.
### P124 - Fig. U05
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** No original caption supplied.
**Purpose:** Country-specific multiple-imputation robustness results. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 19.
[PNG](<08_Robustness_and_Validation/P124-FigU05-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=642>) · Original paper and literature note are not hosted.
### P124 - Fig. U06
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** No original caption supplied.
**Purpose:** Country-specific multiple-imputation robustness results. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 20.
[PNG](<08_Robustness_and_Validation/P124-FigU06-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=643>) · Original paper and literature note are not hosted.
### P124 - Fig. U07
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** No original caption supplied.
**Purpose:** Subgroup robustness analysis using multiple imputation. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 21.
[PNG](<08_Robustness_and_Validation/P124-FigU07-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=644>) · Original paper and literature note are not hosted.
### P124 - Fig. U08
**Paper:** Professional learning communities and teacher outcomes. A cross-national analysis
**Original caption:** No original caption supplied.
**Purpose:** Pooled analysis with teacher and school covariates. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 22.
[PNG](<08_Robustness_and_Validation/P124-FigU08-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=645>) · Original paper and literature note are not hosted.
### P134 - Fig. 7
**Paper:** Separating prior knowledge from acquired knowledge: An individual differences analysis of PISA - learning in the digital world
**Original caption:** Fig. 7. Predicting the Residual with Use of Worked Examples and Intelligence.
**Purpose:** External correlates validate interpretation of the residual as learning. **Visual forms:** Path diagram.
**Physical source pages:** 9.
[PNG](<08_Robustness_and_Validation/P134-Fig07-Predicting%20the%20Residual%20with%20Use%20of%20Worked%20Examples%20and%20Intelligence.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=646>) · Original paper and literature note are not hosted.
### P143 - Fig. 6
**Paper:** Adapting Fairness Interventions to Missing Values
**Original caption:** Figure 6: Comparison of adaptive fairness-intervention algorithms on the synthetic dataset, using DispMistreatment (left) and FairProjection (right). Error bars depict the standard error of 5 runs with different train-test splits.
**Purpose:** Synthetic counterexample checks the adaptive classifiers under a controlled design. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 21.
[PNG](<08_Robustness_and_Validation/P143-Fig06-Comparison%20of%20adaptive%20fairness-intervention%20algorithms%20on%20the%20synthetic%20dataset,%20using%20DispMistreatment%20(left)%20and%20FairProjection%20(right).%20Error%20bars%20depict%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=647>) · Original paper and literature note are not hosted.
### P143 - Fig. 7
**Paper:** Adapting Fairness Interventions to Missing Values
**Original caption:** Figure 7: Comparison of adaptive fairness-intervention algorithms on COMPAS and Adult using DispMistreatment. For each of COMPAS and Adult, we generate three datasets with missing values corresponding to MCAR, MAR, and MNAR mechanisms (Appendix D.2). Error bars depict the standard error of 10 runs with different train-test splits.
**Purpose:** Additional fairness-intervention algorithm checks results across missingness settings. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 21.
[PNG](<08_Robustness_and_Validation/P143-Fig07-Comparison%20of%20adaptive%20fairness-intervention%20algorithms%20on%20COMPAS%20and%20Adult%20using%20DispMistreatment.%20For%20each%20of%20COMPAS%20and%20Adult,%20we%20generate%20three%20datasets%20with....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=648>) · Original paper and literature note are not hosted.
### P143 - Fig. 8
**Paper:** Adapting Fairness Interventions to Missing Values
**Original caption:** Figure 8: Comparison of adding missing indicators (Section 4.1) with baseline impute-then-classify on HSLS, using Reduction, EqOdds, ROC, and Leveraging. Error bars depict the standard error of 15 runs with different train-test splits.
**Purpose:** Additional baseline comparisons test method performance beyond primary experiments. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 22.
[PNG](<08_Robustness_and_Validation/P143-Fig08-Comparison%20of%20adding%20missing%20indicators%20(Section%204.1)%20with%20baseline%20impute-then-classify%20on%20HSLS,%20using%20Reduction,%20EqOdds,%20ROC,%20and%20Leveraging.%20Error%20bars%20depict%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=649>) · Original paper and literature note are not hosted.
### P143 - Fig. 9
**Paper:** Adapting Fairness Interventions to Missing Values
**Original caption:** Figure 9: Results for FairMissBag on COMPAS using FairProjection. Using COMPAS, we generate three datasets with missing values corresponding to MCAR, MAR and MNAR mechanisms (Section D.2). Fairness-accuracy curves for mean, KNN, and iterative imputation are shown. Error bars depict the standard error of 5 runs with different train-test splits.
**Purpose:** Imputation-method sensitivity checks ensemble performance. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 22.
[PNG](<08_Robustness_and_Validation/P143-Fig09-Results%20for%20FairMissBag%20on%20COMPAS%20using%20FairProjection.%20Using%20COMPAS,%20we%20generate%20three%20datasets%20with%20missing%20values%20corresponding%20to%20MCAR,%20MAR%20and%20MNAR%20mechanisms....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=650>) · Original paper and literature note are not hosted.
### P144 - Fig. C.2
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure C.2: Robustness of Diffusion Gaps to Selection into Survey Notes: This figure reports diffusion gaps estimated on the full sample and by randomized response-incentive tier (no incentive, $5, and $10/$20). The outcome is the standardized AI integration index, which aggregates seven standardized survey pillars: written generative AI policy and topics explicitly covered; student use actively encouraged for learning, including studying and homework assistance, with clear rules; teacher use of generative AI tools for teachingrelated tasks by June 2022, 2023, 2024, and 2025; principal use for offloading administrative busywork, streamlining data analysis/reporting, designing professional development, and prepping meetings/presentations; expected extensive or embedded integration three years ahead; a license or contract for any AI-powered tutoring platform; and staff generative AI professional development by June 2025 plus at least 25% of instructional staff completing at least two hours of such professional development. Each reported estimate corresponds to the coefficient on economic disadvantage/FRPL share (Panel a), charter indicator (Panel b), or private indicator (Panel c), estimated in a regression of the standardized AI index on the focal measure, state indicators, racial demographics, missing indicators for schools with missing demographics, and log neighborhood income for the sector panels. All regressions are IPW-weighted, and heteroskedasticity-robust standard errors are used to construct the reported 95 percent confidence intervals. The reported p-value tests the joint null that the focal coefficient is equal across incentive tiers.
**Purpose:** Alternative index, response-selection or Heckman-correction robustness check. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 44.
[PNG](<08_Robustness_and_Validation/P144-FigC.2-Robustness%20of%20Diffusion%20Gaps%20to%20Selection%20into%20Survey%20Notes%20This%20figure%20reports%20diffusion%20gaps%20estimated%20on%20the%20full%20sample%20and%20by%20randomized%20response-incentive....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=651>) · Original paper and literature note are not hosted.
### P144 - Fig. C.3
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure C.3: Robustness to Selection into Survey Responses - Alternative Measures of Disadvantage Notes: This figure replicates Figure C.2 using two alternative measures of school disadvantage. The outcome is the same standardized AI integration index, aggregating written generative AI policy and topics explicitly covered; student use actively encouraged for learning, including studying and homework assistance, with clear rules; teacher use of generative AI tools for teaching-related tasks by June 2022, 2023, 2024, and 2025; principal use for offloading administrative busywork, streamlining data analysis/reporting, designing professional development, and prepping meetings/presentations; expected extensive or embedded integration three years ahead; a license or contract for any AI-powered tutoring platform; and staff generative AI professional development by June 2025 plus at least 25% of instructional staff completing at least two hours of such professional development. Panel (a) reports the coefficient on standardized SEDA average test scores (1 SD) across the full sample and randomized response-incentive tiers, estimated on the public-school sample with controls for racial demographics, missingness indicators, and state fixed effects. Panel (b) reports the coefficient on log median neighborhood household income across the same displayed rows, estimated on the full sample of public and private schools with the same controls. Each estimate is IPW-weighted with heteroskedasticity-robust standard errors used to construct the reported 95 percent confidence intervals. The reported p-value tests the joint null that the coefficient is equal across incentive tiers; failure to reject this null suggests that the diffusion gaps are robust to selection into survey response generated by the randomized incentive arms.
**Purpose:** Alternative index, response-selection or Heckman-correction robustness check. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 45.
[PNG](<08_Robustness_and_Validation/P144-FigC.3-Robustness%20to%20Selection%20into%20Survey%20Responses%20-%20Alternative%20Measures%20of%20Disadvantage%20Notes%20This%20figure%20replicates%20Figure%20C.2%20using%20two%20alternative%20measures%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=652>) · Original paper and literature note are not hosted.
### P144 - Fig. C.4
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure C.4: Diffusion Gaps Using a PCA-Based Index Notes: The PCA index is the first principal component of the seven standardized baseline components, oriented to be positively correlated with the equal-weight baseline index. The components are written generative AI policy/topic depth; student use actively encouraged for learning with clear rules; teacher use of generative AI tools for teachingrelated tasks by June 2022–2025; principal use for administrative busywork, data analysis/reporting, professionaldevelopment design, and meetings/presentations; expected extensive or embedded integration three years ahead; an AI-powered tutoring platform license or contract; and staff generative AI professional development by June 2025 plus at least 25% of instructional staff completing at least two hours of such professional development. PC1 explains 32% of component variance and correlates 1.00 with the baseline index. Specifications, samples, and inference match Figure C.5.
**Purpose:** Alternative index, response-selection or Heckman-correction robustness check. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 46.
[PNG](<08_Robustness_and_Validation/P144-FigC.4-Diffusion%20Gaps%20Using%20a%20PCA-Based%20Index%20Notes%20The%20PCA%20index%20is%20the%20first%20principal%20component%20of%20the%20seven%20standardized%20baseline%20components,%20oriented%20to%20be%20positively....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=653>) · Original paper and literature note are not hosted.
### P144 - Fig. C.5
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure C.5: Robustness of Diffusion Gaps to Leave-One-Component-Out Indexes Notes: Each row corresponds to a re-estimation of the three headline diffusion gaps using an alternative AI integration index that drops one of the seven baseline components: written generative AI policy/topic depth; student use actively encouraged for learning with clear rules; teacher use of generative AI tools for teaching-related tasks by June 2022–2025; principal use for administrative busywork, data analysis/reporting, professional-development design, and meetings/presentations; expected extensive or embedded integration three years ahead; an AI-powered tutoring platform license or contract; and staff generative AI professional development by June 2025 plus at least 25% of instructional staff completing at least two hours of such professional development. The disadvantage gap is the coefficient on standardized FRPL share (1 SD) on the common public-school sample with missingness controls; charter and private gaps are estimated using the adjusted specifications from Figure 4 (state fixed effects, log neighborhood income, racial demographics, and missingness controls). All regressions are IPW-weighted with heteroskedasticityrobust 95% confidence intervals.
**Purpose:** Alternative index, response-selection or Heckman-correction robustness check. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 47.
[PNG](<08_Robustness_and_Validation/P144-FigC.5-Robustness%20of%20Diffusion%20Gaps%20to%20Leave-One-Component-Out%20Indexes%20Notes%20Each%20row%20corresponds%20to%20a%20re-estimation%20of%20the%20three%20headline%20diffusion%20gaps%20using%20an....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=654>) · Original paper and literature note are not hosted.
### P144 - Fig. C.6
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure C.6: Diffusion Gaps for Current-Use vs. Institutional-Support Sub-Indexes Notes: The current-use index combines the student-encouragement, teacher-adoption, and principal-use components: student use actively encouraged for learning, including studying and homework assistance, with clear rules; teacher use of generative AI tools for teaching-related tasks by June 2022–2025; and principal use for offloading administrative busywork, streamlining data analysis/reporting, designing professional development, and prepping meetings/presentations. The institutional-support index combines the written-policy, AI-tutoring-license, and professional-development components: a written policy specifically addressing generative AI use plus topics explicitly covered in that policy; a license or contract for any AI-powered tutoring platform; and staff generative AI professional development by June 2025 plus at least 25% of instructional staff completing at least two hours of such professional development. The expected-future-integration component is included only in the baseline all-seven index. Each index is standardized using IPW weights. Specifications, samples, and inference match Figure C.5.
**Purpose:** Alternative index, response-selection or Heckman-correction robustness check. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 48.
[PNG](<08_Robustness_and_Validation/P144-FigC.6-Diffusion%20Gaps%20for%20Current-Use%20vs.%20Institutional-Support%20Sub-Indexes%20Notes%20The%20current-use%20index%20combines%20the%20student-encouragement,%20teacher-adoption,%20and....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=655>) · Original paper and literature note are not hosted.
### P144 - Fig. E.1
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure E.1: Headline Diffusion Gaps with Heckman Selection Correction Notes: This figure reports the three headline gap estimates with and without the Heckman selection correction. The FRPL disadvantage row is the coefficient on standardized FRPL share in the common public-school sample, controlling for FRPL missingness. The charter row is the coefficient on a charter-school indicator among public schools, and the private row is the coefficient on a private-school indicator among non-charter schools; both sector specifications control for log neighborhood income, income missingness, racial-composition controls and missingness indicators, and state fixed effects. The first-stage probit is fit on the sample used in Tables A.2 and A.3, regressing a respondent indicator on cleaned sample covariates: FRPL, neighborhood income, neighborhood White, Black, and Asian shares, a private-school indicator, and randomized incentive indicators for $5, $10, and $20 offers, with no incentive as the omitted group. The inverse Mills ratio is computed for each respondent and added as an additional control in the Heckman-corrected specifications. All regressions are IPW-weighted with heteroskedasticity-robust 95% confidence intervals.
**Purpose:** Alternative index, response-selection or Heckman-correction robustness check. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 51.
[PNG](<08_Robustness_and_Validation/P144-FigE.1-Headline%20Diffusion%20Gaps%20with%20Heckman%20Selection%20Correction%20Notes%20This%20figure%20reports%20the%20three%20headline%20gap%20estimates%20with%20and%20without%20the%20Heckman%20selection....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=656>) · Original paper and literature note are not hosted.
### P144 - Fig. E.2
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure E.2: Mechanism Coefficients with Heckman Selection Correction Notes: This figure reports mechanism coefficients with and without the Heckman selection correction. The outcome is the standardized AI integration index. Surveyed mechanism rows include positive student-learning-impact beliefs; teacher champions as the factor that most enabled current AI efforts; principal authority to purchase new instructional software; discretionary technology budget of at least $5,000; redirecting previously allocated funds toward AI initiatives and AI-related tools as easy or very easy; ESSER use for AI-related purchases; and cost, student misuse, teacher skill gap, and policy restriction barriers selected from the up-to-two-barriers item. Baseline specifications control for log neighborhood income, income missingness, and state fixed effects. Heckman-corrected specifications additionally include the inverse Mills ratio from the first-stage probit described in Figure E.1. All regressions are IPW-weighted with heteroskedasticity-robust 95% confidence intervals.
**Purpose:** Alternative index, response-selection or Heckman-correction robustness check. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 52.
[PNG](<08_Robustness_and_Validation/P144-FigE.2-Mechanism%20Coefficients%20with%20Heckman%20Selection%20Correction%20Notes%20This%20figure%20reports%20mechanism%20coefficients%20with%20and%20without%20the%20Heckman%20selection%20correction.%20The....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=657>) · Original paper and literature note are not hosted.
### P144 - Fig. E.3
**Paper:** AI Diffusion Gaps: Unequal Integration of AI Across K-12 Schools
**Original caption:** Figure E.3: Gap Stability Across Mechanism Specifications — Heckman Correction Notes: This figure reports gap stability across sequential mechanism specifications, comparing baseline and Heckmancorrected coefficients. The disadvantage estimate is the coefficient on standardized FRPL share among public schools; the charter estimate is the coefficient on a charter school indicator among public schools; and the private estimate is the coefficient on a private school indicator among non-charter schools. Sequential specifications cumulatively add positive-impact beliefs, leadership controls, budget controls, ESSER use, and barrier controls. The disadvantage specifications control for FRPL missingness; the sector specifications control for log neighborhood income and income missingness. Heckman-corrected specifications additionally include the inverse Mills ratio from the first-stage probit described in Figure E.1. All regressions are IPW-weighted with state fixed effects and heteroskedasticity-robust 95% confidence intervals.
**Purpose:** Alternative index, response-selection or Heckman-correction robustness check. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 53.
[PNG](<08_Robustness_and_Validation/P144-FigE.3-Gap%20Stability%20Across%20Mechanism%20Specifications%20-%20Heckman%20Correction%20Notes%20This%20figure%20reports%20gap%20stability%20across%20sequential%20mechanism%20specifications,%20comparing....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=658>) · Original paper and literature note are not hosted.
### P146 - Fig. 2
**Paper:** A meta-analysis on the effect of technology on the achievement of less advantaged students
**Original caption:** Fig. 2. Funnel plot.
**Purpose:** Publication-bias funnel diagnostic. **Visual forms:** Scatter.
**Physical source pages:** 12.
[PNG](<08_Robustness_and_Validation/P146-Fig02-Funnel%20plot.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=659>) · Original paper and literature note are not hosted.
### P146 - Fig. 3
**Paper:** A meta-analysis on the effect of technology on the achievement of less advantaged students
**Original caption:** Fig. 3. Doi plot.
**Purpose:** Doi-plot small-study/publication-bias diagnostic. **Visual forms:** Scatter.
**Physical source pages:** 12.
[PNG](<08_Robustness_and_Validation/P146-Fig03-Doi%20plot.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=660>) · Original paper and literature note are not hosted.
### P149 - Fig. 2
**Paper:** A systematic review and meta-analysis of the evidence on learning during the COVID-19 pandemic
**Original caption:** Fig. 2 | Risk of bias and publication bias. a, Domain-specific and overall distribution of studies of COVID-19 learning deficits by risk of bias rating using ROBINS-I, including studies rated to be at critical risk of bias (n = 19 out of a total of n = 61 studies shown in this figure). In line with ROBINS-I guidance, studies rated to be at critical risk of bias were excluded from all analyses and other figures in this article and in the Supplementary Information (including b). b, z curve: distribution of the z scores of all estimates included in the metaanalysis (n = 291) to test for publication bias. The dotted line indicates z = 1.96 (P = 0.050), the conventional threshold for statistical significance. The overlaid curve shows a normal distribution. The absence of a spike in the distribution of the z scores just above the threshold for statistical significance and the absence of a slump just below it indicate the absence of evidence for publication bias.
**Purpose:** Assesses study risk of bias and publication bias. **Visual forms:** Composite, Stacked bar chart, Histogram.
**Physical source pages:** 3.
[PNG](<08_Robustness_and_Validation/P149-Fig02-Risk%20of%20bias%20and%20publication%20bias.%20a,%20Domain-specific%20and%20overall%20distribution%20of%20studies%20of%20COVID-19%20learning%20deficits%20by%20risk%20of%20bias%20rating%20using%20ROBINS-I,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=661>) · Original paper and literature note are not hosted.
### P158 - Fig. 5
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Fig. 5 Funnel plot of standard error by Hedges’ g
**Purpose:** Publication-bias diagnostic. **Visual forms:** Funnel plot.
**Physical source pages:** 23.
[PNG](<08_Robustness_and_Validation/P158-Fig05-Funnel%20plot%20of%20standard%20error%20by%20Hedges%E2%80%99%20g.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=662>) · Original paper and literature note are not hosted.
### P158 - Table 6
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Table 6 Risk of bias ratings (RoB 2) for randomized studies. RCT = randomized control trial; cRCT = clustered randomized control trial
**Purpose:** Structured domain-level quality appraisal of randomized evidence. **Visual forms:** Complex table, Risk-of-bias matrix.
**Physical source pages:** 89, 90, 91.
[PNG](<08_Robustness_and_Validation/P158-Table06-Risk%20of%20bias%20ratings%20(RoB%202)%20for%20randomized%20studies.%20RCT%20=%20randomized%20control%20trial;%20cRCT%20=%20clustered%20randomized%20control%20trial.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=663>) · Original paper and literature note are not hosted.
### P158 - Table 7
**Paper:** Evidence of impact and interpretational limits of generative AI in STEM education: a systematic review and meta-analysis on cognitive learning outcomes
**Original caption:** Table 7 Risk of bias (ROBINS-I) for non-randomized studies. QE = Quasi-experimental; CB = Controlled before-after; CO = Cohort; CE = Controlled experiment; DBR = Design-based research; MMS QE = Mixed-methods quasi-experimental; ni = no information. D4 = Deviations from intended interventions
**Purpose:** Structured domain-level quality appraisal of non-randomized evidence. **Visual forms:** Complex table, Risk-of-bias matrix.
**Physical source pages:** 92, 93, 94.
[PNG](<08_Robustness_and_Validation/P158-Table07-Risk%20of%20bias%20(ROBINS-I)%20for%20non-randomized%20studies.%20QE%20=%20Quasi-experimental;%20CB%20=%20Controlled%20before-after;%20CO%20=%20Cohort;%20CE%20=%20Controlled%20experiment;%20DBR%20=....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=664>) · Original paper and literature note are not hosted.
### P159 - Fig. A4
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure A4: Test-Score Effects Excluding One Time Slot at a Time Notes: This figure re-estimates the test-score ITT for Session One (Panel A) and Session Two (Panel B) excluding one of the eight time slots (the student’s Session One time slot, the randomization stratum) at a time. The hollow marker keeps all slots and reproduces the estimates in Table 4 (the dashed line marks that benchmark). Each estimate re-selects double-lasso controls on its subsample (Belloni et al., 2014) and includes strata fixed effects. Vertical bars represent 95 percent confidence intervals from heteroskedasticityrobust standard errors. Excluding a slot leaves 173–190 students in Session One and 171–185 in Session Two.
**Purpose:** Shows test-score estimates after excluding each session time slot in turn. **Visual forms:** Line/profile plot.
**Physical source pages:** 46.
[PNG](<08_Robustness_and_Validation/P159-FigA4-Test-Score%20Effects%20Excluding%20One%20Time%20Slot%20at%20a%20Time%20Notes%20This%20figure%20re-estimates%20the%20test-score%20ITT%20for%20Session%20One%20(Panel%20A)%20and%20Session%20Two%20(Panel%20B)%20excluding....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=665>) · Original paper and literature note are not hosted.
### P160 - Fig. B.1
**Paper:** Explaining the learning decline in PISA 2025 by socioeconomic quartiles through methodological changes
**Original caption:** Figure B.1. Science: the fan of the published HOMEPOS disappears once item parameters are fixed. Change between 2022 and 2025 in the mean science score of each quartile of the index. Colour identifies the version: blue denotes the published index (OECD) and orange the harmonized index (this paper). Solid lines show ESCS and dashed lines HOMEPOS. The harmonized HOMEPOS uses a 2PL/GPCM model calibrated once on both cycles with the 16 common items, whereas the published index changed items between cycles; the harmonized ESCS combines the three harmonized components with the OECD’s rule for the composite. Grey lines show parental occupation (HISEI, solid) and parental education (PARED, dotted) on their harmonized metric, which have no published counterpart. Quartiles hold approximately 25% of the student weight each and are computed within each system and cycle; values are simple averages of 36 OECD systems. Vertical error bars show 95% confidence intervals for each quartile’s change in mean score. Standard errors combine sampling variance from the 80 BRR replicates and imputation variance across the ten plausible values, plus the OECD cross-cycle linking variance.
**Purpose:** Compares published versus harmonized socioeconomic measures and the resulting achievement trend; appendix displays validate across alternative domains. **Visual forms:** Line plot, Point and interval plot.
**Physical source pages:** 29.
[PNG](<08_Robustness_and_Validation/P160-FigB.1-Science%20the%20fan%20of%20the%20published%20HOMEPOS%20disappears%20once%20item%20parameters%20are%20fixed.%20Change%20between%202022%20and%202025%20in%20the%20mean%20science%20score%20of%20each%20quartile%20of%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=666>) · Original paper and literature note are not hosted.
### P160 - Fig. B.2
**Paper:** Explaining the learning decline in PISA 2025 by socioeconomic quartiles through methodological changes
**Original caption:** Figure B.2. Level of the socioeconomic gap in science, 2015–2025: published versus harmonized indices. Difference in mean science score between the top and bottom quartiles (Q4 − Q1) of ESCS (solid lines) and of HOMEPOS (dashed lines) Orange denotes the harmonized four-cycle index (this paper), which uses the 13 possessions items whose concept exists in every cycle with one calibration for the four cycles, parental education mapped to a common scale, and the OECD’s rule for the composite; blue denotes the index as published by the OECD. Quartiles hold approximately 25% of the student weight each and are computed within each system and cycle; values are simple averages of the 34 OECD systems present in all four cycles. Vertical error bars show 95% confidence intervals for each estimated Q4 − Q1 gap. Standard errors combine sampling variance from the 80 BRR replicates and imputation variance across the ten plausible values.
**Purpose:** Compares published versus harmonized socioeconomic measures and the resulting achievement trend; appendix displays validate across alternative domains. **Visual forms:** Line plot, Point and interval plot.
**Physical source pages:** 30.
[PNG](<08_Robustness_and_Validation/P160-FigB.2-Level%20of%20the%20socioeconomic%20gap%20in%20science,%202015-2025%20published%20versus%20harmonized%20indices.%20Difference%20in%20mean%20science%20score%20between%20the%20top%20and%20bottom%20quartiles%20(Q4....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=667>) · Original paper and literature note are not hosted.
### P160 - Fig. B.3
**Paper:** Explaining the learning decline in PISA 2025 by socioeconomic quartiles through methodological changes
**Original caption:** Figure B.3. Change 2022–2025 in mean science score by ESCS quartile, by system. Each panel is one education system, identified by its three-letter code; panels are ordered by the change in the Q4 − Q1 gap under the published ESCS, from the largest narrowing (top left) to the largest widening (bottom right). Orange lines show harmonized ESCS (this paper), which rescales the household possessions component on the 16 items present in both cycles with a single joint calibration and follows the OECD’s rule for the composite; blue lines show ESCS as published by the OECD. Points are the change in the mean score of each equal-weight quartile, computed within system and cycle; the axes are shared across panels. Standard errors of the change in the gap by system are given in Table B.7.
**Purpose:** Compares published versus harmonized socioeconomic measures and the resulting achievement trend; appendix displays validate across alternative domains. **Visual forms:** Composite, Line plot.
**Physical source pages:** 31.
[PNG](<08_Robustness_and_Validation/P160-FigB.3-Change%202022-2025%20in%20mean%20science%20score%20by%20ESCS%20quartile,%20by%20system.%20Each%20panel%20is%20one%20education%20system,%20identified%20by%20its%20three-letter%20code;%20panels%20are%20ordered%20by....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=668>) · Original paper and literature note are not hosted.
### P160 - Fig. C.1
**Paper:** Explaining the learning decline in PISA 2025 by socioeconomic quartiles through methodological changes
**Original caption:** Figure C.1. Reading: the fan of the published HOMEPOS disappears once item parameters are fixed. Change between 2022 and 2025 in the mean reading score of each quartile of the index. Colour identifies the version: blue denotes the published index (OECD) and orange the harmonized index (this paper). Solid lines show ESCS and dashed lines HOMEPOS. The harmonized HOMEPOS uses a 2PL/GPCM model calibrated once on both cycles with the 16 common items, whereas the published index changed items between cycles; the harmonized ESCS combines the three harmonized components with the OECD’s rule for the composite. Grey lines show parental occupation (HISEI, solid) and parental education (PARED, dotted) on their harmonized metric, which have no published counterpart. Quartiles hold approximately 25% of the student weight each and are computed within each system and cycle; values are simple averages of 36 OECD systems. Vertical error bars show 95% confidence intervals for each quartile’s change in mean score. Standard errors combine sampling variance from the 80 BRR replicates and imputation variance across the ten plausible values, plus the OECD cross-cycle linking variance.
**Purpose:** Compares published versus harmonized socioeconomic measures and the resulting achievement trend; appendix displays validate across alternative domains. **Visual forms:** Line plot, Point and interval plot.
**Physical source pages:** 35.
[PNG](<08_Robustness_and_Validation/P160-FigC.1-Reading%20the%20fan%20of%20the%20published%20HOMEPOS%20disappears%20once%20item%20parameters%20are%20fixed.%20Change%20between%202022%20and%202025%20in%20the%20mean%20reading%20score%20of%20each%20quartile%20of%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=669>) · Original paper and literature note are not hosted.
### P160 - Fig. C.2
**Paper:** Explaining the learning decline in PISA 2025 by socioeconomic quartiles through methodological changes
**Original caption:** Figure C.2. Level of the socioeconomic gap in reading, 2015–2025: published versus harmonized indices. Difference in mean reading score between the top and bottom quartiles (Q4 − Q1) of ESCS (solid lines) and of HOMEPOS (dashed lines) Orange denotes the harmonized four-cycle index (this paper), which uses the 13 possessions items whose concept exists in every cycle with one calibration for the four cycles, parental education mapped to a common scale, and the OECD’s rule for the composite; blue denotes the index as published by the OECD. Quartiles hold approximately 25% of the student weight each and are computed within each system and cycle; values are simple averages of the 34 OECD systems present in all four cycles. Vertical error bars show 95% confidence intervals for each estimated Q4 − Q1 gap. Standard errors combine sampling variance from the 80 BRR replicates and imputation variance across the ten plausible values.
**Purpose:** Compares published versus harmonized socioeconomic measures and the resulting achievement trend; appendix displays validate across alternative domains. **Visual forms:** Line plot, Point and interval plot.
**Physical source pages:** 36.
[PNG](<08_Robustness_and_Validation/P160-FigC.2-Level%20of%20the%20socioeconomic%20gap%20in%20reading,%202015-2025%20published%20versus%20harmonized%20indices.%20Difference%20in%20mean%20reading%20score%20between%20the%20top%20and%20bottom%20quartiles%20(Q4....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=670>) · Original paper and literature note are not hosted.
### P160 - Fig. C.3
**Paper:** Explaining the learning decline in PISA 2025 by socioeconomic quartiles through methodological changes
**Original caption:** Figure C.3. Change 2022–2025 in mean reading score by ESCS quartile, by system. Each panel is one education system, identified by its three-letter code; panels are ordered by the change in the Q4 − Q1 gap under the published ESCS, from the largest narrowing (top left) to the largest widening (bottom right). Orange lines show harmonized ESCS (this paper), which rescales the household possessions component on the 16 items present in both cycles with a single joint calibration and follows the OECD’s rule for the composite; blue lines show ESCS as published by the OECD. Points are the change in the mean score of each equal-weight quartile, computed within system and cycle; the axes are shared across panels. Standard errors of the change in the gap by system are given in Table C.7.
**Purpose:** Compares published versus harmonized socioeconomic measures and the resulting achievement trend; appendix displays validate across alternative domains. **Visual forms:** Composite, Line plot.
**Physical source pages:** 37.
[PNG](<08_Robustness_and_Validation/P160-FigC.3-Change%202022-2025%20in%20mean%20reading%20score%20by%20ESCS%20quartile,%20by%20system.%20Each%20panel%20is%20one%20education%20system,%20identified%20by%20its%20three-letter%20code;%20panels%20are%20ordered%20by....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=671>) · Original paper and literature note are not hosted.
### P166 - Fig. 6
**Paper:** Imputation and missing indicators for handling missing data in the development and deployment of clinical prediction models: A simulation study
**Original caption:** Figure 6. Predictive performance of imputation methods, model forms and imputation/validation strategies. DY = Y included in the imputation model at development, DȲ = Y omitted from the imputation model at development, VY = Y included in the imputation model at validation, VȲ = Y omitted from the imputation model at validation, DA = complete data available at model development, VA = complete data available at validation, VYI = Y imputed at validation.
**Purpose:** Real clinical-data validation/application of simulation-derived imputation comparisons. **Visual forms:** Forest/coefficient plot, Composite.
**Physical source pages:** 13.
[PNG](<08_Robustness_and_Validation/P166-Fig06-Predictive%20performance%20of%20imputation%20methods,%20model%20forms%20and%20imputation%20validation%20strategies.%20DY%20=%20Y%20included%20in%20the%20imputation%20model%20at%20development,%20D%C8%B2%20=%20Y....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=672>) · Original paper and literature note are not hosted.
### P173 - Fig. 1
**Paper:** jomo: A Flexible Package for Two-level Joint Modelling Multiple Imputation
**Original caption:** Figure 1: MCMC chain for βe,0 (left panel) and ω2 e,1,1 (right panel).
**Purpose:** MCMC chain diagnostics assess convergence. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 16.
[PNG](<08_Robustness_and_Validation/P173-Fig01-MCMC%20chain%20for%20%CE%B2e,0%20(left%20panel)%20and%20%CF%892%20e,1,1%20(right%20panel).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=673>) · Original paper and literature note are not hosted.
### P175 - Fig. Extended Data 4
**Paper:** Machine learning and phone data can improve targeting of humanitarian aid
**Original caption:** Extended Data Fig. 4 | Targeting performance at different levels of program coverage. Top figures (a and b) show performance for the rural Novissi program, evaluated using 2020 phone survey. Bottom figures (c and d) correspond to the hypothetical national program, evaluated using the 2018–2019 field survey. ROC curves on left (a and c) indicate the true positive and false positive rates at different targeting thresholds. Coverage vs. Recall figures on right (b and d) show how precision and recall vary as the percentage of the population receiving benefits increases, i.e., they indicate the precision and recall for reaching the poorest k% of the population in programs that target the poorest k%. (Precision and recall are thus the same for each value of k by construction; see Methods, ‘Measures of targeting quality’).
**Purpose:** Coverage-threshold sensitivity compares targeting methods under rural and national evaluations. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 21.
[PNG](<08_Robustness_and_Validation/P175-FigExtended%20Data%204-Extended%20Data%20Fig.%204%20Targeting%20performance%20at%20different%20levels%20of%20program%20coverage.%20Top%20figures%20(a%20and%20b)%20show%20performance%20for%20the%20rural%20Novissi....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=674>) · Original paper and literature note are not hosted.
### P177 - Fig. 3
**Paper:** Missing data and prediction: the pattern submodel
**Original caption:** Fig. 3. The continuous SPS measurement was dichotomized at the median, and then used as a binary outcome. The covariates included in the SPS prediction model include partial pressure of oxygen in the arterial blood (pafi), mean blood pressure (meanbp), white blood count (wblc), albumin (alb), APACHE III respiration score (resp), temperature (temp), heart rate per minute (hrt), bilirubin (bili), creatinine (crea), and sodium (sod). There are 23 patterns present in the SUPPORT data, and missing covariates are denoted with “X”. N is the total number of subjects in each missing data pattern. Pattern submodels (PS), relaxed lasso PS, multiple imputation with missingness indicators (MIMI), complete case submodels (CCS), relaxed Lasso CCS, and traditional multiple imputation (MI) methods are all compared. The top two figures are the brier score (BS; unweighted pattern specific BS and weighted pattern specific BS), and the bottom two figures are the log-score (LS; unweighted pattern specific LS and weighted pattern specific LS). The prediction measures are cross-validated (10-fold), with nested cross-validation of the submodels.
**Purpose:** Alternative binary-outcome analysis with cross-validated loss measures. **Visual forms:** Scatter, Complex table, Composite.
**Physical source pages:** 15.
[PNG](<08_Robustness_and_Validation/P177-Fig03-The%20continuous%20SPS%20measurement%20was%20dichotomized%20at%20the%20median,%20and%20then%20used%20as%20a%20binary%20outcome.%20The%20covariates%20included%20in%20the%20SPS%20prediction%20model%20include%20partial....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=675>) · Original paper and literature note are not hosted.
### P178 - Fig. 1
**Paper:** Multilevel models with multivariate mixed response types
**Original caption:** Figure 1 Chain for 5000 iterations for slope variance parameter
**Purpose:** Slope-variance MCMC diagnostic. **Visual forms:** Line/profile plot.
**Physical source pages:** 16.
[PNG](<08_Robustness_and_Validation/P178-Fig01-Chain%20for%205000%20iterations%20for%20slope%20variance%20parameter.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=676>) · Original paper and literature note are not hosted.
### P178 - Fig. 3
**Paper:** Multilevel models with multivariate mixed response types
**Original caption:** Figure 3 MCMC chain of 5000 samples for threshold parameter 1
**Purpose:** Threshold-parameter MCMC diagnostic. **Visual forms:** Line/profile plot.
**Physical source pages:** 20.
[PNG](<08_Robustness_and_Validation/P178-Fig03-MCMC%20chain%20of%205000%20samples%20for%20threshold%20parameter%201.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=677>) · Original paper and literature note are not hosted.
### P178 - Fig. 4
**Paper:** Multilevel models with multivariate mixed response types
**Original caption:** Figure 4 MCMC chain of 5000 samples for level 2 variance
**Purpose:** Level-two variance MCMC diagnostic. **Visual forms:** Line/profile plot.
**Physical source pages:** 21.
[PNG](<08_Robustness_and_Validation/P178-Fig04-MCMC%20chain%20of%205000%20samples%20for%20level%202%20variance.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=678>) · Original paper and literature note are not hosted.
### P182 - Fig. 6
**Paper:** Patterns of item nonresponse behaviour to survey questionnaires are systematic and associated with genetic loci
**Original caption:** Fig. 6 | GWAS results for lead SNPs associated with FI with and without Heckman correction for missingness in N = 83,677 UKB participants in a follow-up FI survey. Deming regression of the observed betas in the two GWAS (solid blue line) indicates that betas from the Heckman corrected GWAS are, on average, attenuated downward from being equal to the uncorrected GWAS (dashed red line). Grey error bars reflect ±1 s.e. on the estimated betas for each SNP. Lead SNPs for GWAS of FI identified using FUMA88. Deming regression performed with the R package ‘deming’101.
**Purpose:** Compares genetic association estimates before and after a missingness correction as a proof-of-concept validation. **Visual forms:** Scatter.
**Physical source pages:** 11.
[PNG](<08_Robustness_and_Validation/P182-Fig06-GWAS%20results%20for%20lead%20SNPs%20associated%20with%20FI%20with%20and%20without%20Heckman%20correction%20for%20missingness%20in%20N%20=%2083,677%20UKB%20participants%20in%20a%20follow-up%20FI%20survey.%20Deming....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=679>) · Original paper and literature note are not hosted.
### P185 - Fig. 2
**Paper:** Retrieval Practice Produces More Learning than Elaborative Studying with Concept Mapping
**Original caption:** Fig. 2. Results of Experiment 2. (A and B) show the proportions correct on the final short-answer tests for enumeration and sequence texts, respectively. (C and D) show the proportions correct on the final concept mapping tests for enumeration and sequence texts, respectively. Error bars indicate SEM. Retrieval practice enhanced long-term learning above and beyond elaborative concept mapping by more than one standard deviation on both types of final test (d = 1.07 and 1.01, respectively). (E to H) show the proportion of information subjects predicted they would recall on the final test in each initial learning condition. Students tended to believe that elaborative concept mapping would produce the same or even greater learning than retrieval practice, even though the opposite was true, as shown in (A) to (D).
**Purpose:** Second experiment tests generalization across materials and final-test formats. **Visual forms:** Bar chart, Composite.
**Physical source pages:** 3.
[PNG](<08_Robustness_and_Validation/P185-Fig02-Results%20of%20Experiment%202.%20(A%20and%20B)%20show%20the%20proportions%20correct%20on%20the%20final%20short-answer%20tests%20for%20enumeration%20and%20sequence%20texts,%20respectively.%20(C%20and%20D)%20show%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=680>) · Original paper and literature note are not hosted.
### P188 - Fig. 3
**Paper:** Socioeconomic status and student learning: insights from an umbrella review
**Original caption:** Fig. 3 Funnel plot
**Purpose:** Publication-bias diagnostic for the second-order meta-analysis. **Visual forms:** Scatter.
**Physical source pages:** 33.
[PNG](<08_Robustness_and_Validation/P188-Fig03-Funnel%20plot.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=681>) · Original paper and literature note are not hosted.
### P189 - Fig. 2
**Paper:** Specification curve analysis
**Original caption:** Fig. 2 | Descriptive specification curve. Each dot in the top panel (green area) depicts the marginal effect, estimated at sample means, of a hurricane having a female rather than male name; the dots vertically aligned below (white area) indicate the analytical decisions behind those estimates. A total of 1,728 specifications were estimated; to facilitate visual inspection, the figure depicts the 50 highest and lowest point estimates and a random subset of 200 additional ones, but the inferential statistics for specification curve analysis include all 1,728 specifications. NS, not significant.
**Purpose:** Complete model universe diagnoses robustness to analytical choices. **Visual forms:** Specification curve, Matrix.
**Physical source pages:** 3.
[PNG](<08_Robustness_and_Validation/P189-Fig02-Descriptive%20specification%20curve.%20Each%20dot%20in%20the%20top%20panel%20(green%20area)%20depicts%20the%20marginal%20effect,%20estimated%20at%20sample%20means,%20of%20a%20hurricane%20having%20a%20female%20rather....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=682>) · Original paper and literature note are not hosted.
### P189 - Fig. 3
**Paper:** Specification curve analysis
**Original caption:** Fig. 3 | Observed and expected under-the-null specification curves for the hurricanes and racial discrimination studies. Observed and expected under-the-null specification curves for the hurricanes and racial discrimination studies. The expected curves are based on 500 shuffled samples where the key predictor in each dataset (hurricane and applicant name, respectively) is shuffled. All specifications are estimated on each shuffled sample (1,728 specifications for hurricanes study, 90 for racial discrimination). The resulting estimates for each shuffled dataset are ranked from smallest to largest. The dashed lines depict the 2.5th, 50th and 97.5th percentiles for each of these ranked estimates (for example, the median smallest estimate, the median second smallest estimate and so on). Specification curves under the null are typically not symmetric around zero (see Main text). Blue dots depict the specification curve for the observed data.
**Purpose:** Observed versus joint-null curves assess robustness and inferential validity across specifications. **Visual forms:** Specification curve, Composite.
**Physical source pages:** 5.
[PNG](<08_Robustness_and_Validation/P189-Fig03-Observed%20and%20expected%20under-the-null%20specification%20curves%20for%20the%20hurricanes%20and%20racial%20discrimination%20studies.%20Observed%20and%20expected%20under-the-null%20specification....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=683>) · Original paper and literature note are not hosted.
### P191 - Fig. B.1
**Paper:** Substitute or Complement? Generative AI Use and Student Achievement: Evidence from PISA 2025
**Original caption:** Figure B.1: The level coefficient under alternative definitions of inattentive responding Notes: Coefficient on the standardized chatbot index, in PISA points per standard deviation, with 95 percent intervals, from specifications with socioeconomic status and school fixed effects in which the row’s definition of inattentive responding is either added as a control (filled markers) or used to exclude the flagged students (hollow markers), as the only response-style term. The dashed line is the paper’s own within-school estimate, the last row of Figure 2. The first two rows are references: the specification with socioeconomic status and fixed effects only, and the paper’s own specification. Definitions are in Table B.1; exact estimates, including country fixed effects, in Table B.2. Senate-weighted; standard errors combine 80 replicate weights with the imputation variance across ten plausible values.
**Purpose:** Tests robustness of the level coefficient under alternative inattentive-response definitions. **Visual forms:** Composite, Forest/coefficient plot.
**Physical source pages:** 6.
[PNG](<08_Robustness_and_Validation/P191-App-FigB.1-The%20level%20coefficient%20under%20alternative%20definitions%20of%20inattentive%20responding%20Notes%20Coefficient%20on%20the%20standardized%20chatbot%20index,%20in%20PISA%20points%20per%20standard....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=684>) · Original paper and literature note are not hosted.
### P191 - Fig. B.2
**Paper:** Substitute or Complement? Generative AI Use and Student Achievement: Evidence from PISA 2025
**Original caption:** Figure B.2: The purpose coefficient under alternative definitions of inattentive responding. Notes: Coefficient on the second principal component, learning against producing, from the two components entered together, in PISA points per standard deviation with 95 percent intervals. Specifications and markers as in Figure B.1; the dashed line is the paper’s within-school estimate, the hollow PC2 marker of Figure 4. Exact estimates are in Table B.3.
**Purpose:** Tests robustness of the purpose coefficient under alternative inattentive-response definitions. **Visual forms:** Composite, Forest/coefficient plot.
**Physical source pages:** 7.
[PNG](<08_Robustness_and_Validation/P191-App-FigB.2-The%20purpose%20coefficient%20under%20alternative%20definitions%20of%20inattentive%20responding.%20Notes%20Coefficient%20on%20the%20second%20principal%20component,%20learning%20against....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=685>) · Original paper and literature note are not hosted.
### P191 - Fig. C.1
**Paper:** Substitute or Complement? Generative AI Use and Student Achievement: Evidence from PISA 2025
**Original caption:** Figure C.1: The level and purpose coefficients under alternative control sets. Notes: PISA points per standard deviation, with 95 percent intervals that combine 80 replicate weights with the imputation variance across ten plausible values. Rows add the two control blocks cumulatively to the paper’s specification; within a row, filled markers use country fixed effects and hollow markers school fixed effects, and the labels give the within-school estimates. Panel (a) is the coefficient on the chatbot index entered alone, the level margin of Figure 2; panel (b) the coefficient on the second principal component from the two entered together, the purpose margin of Figure 4. Exact estimates for mathematics, including the four items, are in Table C.1.
**Purpose:** Compares the main level and purpose estimates across successively expanded control sets. **Visual forms:** Composite, Forest/coefficient plot.
**Physical source pages:** 13.
[PNG](<08_Robustness_and_Validation/P191-App-FigC.1-The%20level%20and%20purpose%20coefficients%20under%20alternative%20control%20sets.%20Notes%20PISA%20points%20per%20standard%20deviation,%20with%2095%20percent%20intervals%20that%20combine%2080%20replicate....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=686>) · Original paper and literature note are not hosted.
### P191 - Fig. E.2
**Paper:** Substitute or Complement? Generative AI Use and Student Achievement: Evidence from PISA 2025
**Original caption:** Figure E.2: The chatbot index against the index distributed with the assessment
**Purpose:** Compares the paper’s index against the OECD index and shows the adjusted achievement gradient under each measure. **Visual forms:** Composite, Scatter.
**Physical source pages:** 18.
[PNG](<08_Robustness_and_Validation/P191-App-FigE.2-The%20chatbot%20index%20against%20the%20index%20distributed%20with%20the%20assessment.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=687>) · Original paper and literature note are not hosted.
### P196 - Fig. 3
**Paper:** Universal Basic Incomes vs. Targeted Transfers: Anti-Poverty Programs in Developing Countries
**Original caption:** Figure 3: Predicted vs. actual per-capita consumption in test set data Notes: The figures above plot actual log per-capita monthly consumption against predicted log per-capital monthly consumption for both Indonesia and Peru. The solid vertical and horizontal lines represent eligibility cutoffs for the programs we model here. (In Indonesia, we set the cutoff approximately at the 33rd percentile for consumption; and in Peru, we set the cutoff at the 28th percentile.) The dashed line is a 45˚ line. For legibility, the points plotted above represent a random sample of 10 percent of our full data from Indonesia and 50 percent of our full data from Peru.
**Purpose:** Checks out-of-sample consumption prediction and targeting errors. **Visual forms:** Composite, Scatter.
**Physical source pages:** 31.
[PNG](<08_Robustness_and_Validation/P196-Fig03-Predicted%20vs.%20actual%20per-capita%20consumption%20in%20test%20set%20data%20Notes%20The%20figures%20above%20plot%20actual%20log%20per-capita%20monthly%20consumption%20against%20predicted%20log%20per-capital....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=688>) · Original paper and literature note are not hosted.
### P196 - Fig. 7
**Paper:** Universal Basic Incomes vs. Targeted Transfers: Anti-Poverty Programs in Developing Countries
**Original caption:** Figure 7: Implied Tax Rate vs. Household Income Notes: Figure 7 plots the implied tax rate against actual per-capita household consumption for both Indonesia and Peru. We designated households as benefit recipients if predicted income was less than 1.5x the poverty line (in Indonesia) or less than 1x the poverty line (in Peru). Households above the 95th income percentile were dropped from both country samples. We then performed a local polynomial regression of benefit receipt status on actual income, using a bandwidth of 50,000 rupiah (Indonesia) and 75 soles (Peru). We took the first derivative of this function and multiplied it by (minus) the benefit amount in order to calculate the implied tax rate as a function of household income (−p′(y)b). We then applied the same process after adding a random, normally distributed noise term (with the same standard deviation as our original predicted incomes) to these predicted income values.
**Purpose:** Examines implied tax-rate results when prediction noise is doubled. **Visual forms:** Composite, Line plot.
**Physical source pages:** 35.
[PNG](<08_Robustness_and_Validation/P196-Fig07-Implied%20Tax%20Rate%20vs.%20Household%20Income%20Notes%20Figure%207%20plots%20the%20implied%20tax%20rate%20against%20actual%20per-capita%20household%20consumption%20for%20both%20Indonesia%20and%20Peru.%20We....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=689>) · Original paper and literature note are not hosted.
### P203 - Fig. 3
**Paper:** A Systematic Review and Meta-Analysis of Discrepancies Between Logged and Self-Reported Digital Media Use
**Original caption:** Figure 3. Contour-enhanced funnel plots. The plot depicts the relationship between the observed effect sizes (on the x-axis) and their standard errors (on the y-axis) for comparisons concerning digital media use (A), problematic use (B) and reporting accuracy (C). The vertical lines indicate the estimated summary effect size. The shaded bands represent the significance contours indicated in the legend and each black dot represents an observed effect size. Visual inspection of all three plots does not indicate asymmetry, nor does it indicate evidence of publication bias as there is no obvious overrepresentation of effect sizes in the highlighted significance contours.
**Purpose:** Publication/small-study bias diagnostics. **Visual forms:** Density/histogram/boxplot, Composite.
**Physical source pages:** 26, 40.
[PNG](<08_Robustness_and_Validation/P203-Fig03-Contour-enhanced%20funnel%20plots.%20The%20plot%20depicts%20the%20relationship%20between%20the%20observed%20effect%20sizes%20(on%20the%20x-axis)%20and%20their%20standard%20errors%20(on%20the%20y-axis)%20for....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=964>) · Original paper and literature note are not hosted.
### P203 - Fig. Extended Data 1
**Paper:** A Systematic Review and Meta-Analysis of Discrepancies Between Logged and Self-Reported Digital Media Use
**Original caption:** Extended Data Fig. 1 Reporting accuracy post hoc moderator and subgroup analyses. Note: k: number of separate effect sizes included for the moderator level; R = response ratio; Exp(β) = exponential transformation of metaregression coefficient from a model in which a categorical moderator with two levels was entered as a predictor. F values correspond to the Approximate Hotelling-Zhang with small sample correction omnibus tests for moderators with more than two levels; 95% CI corresponds to the r values for individual moderator levels; p corresponds to the F value for moderators or the subgroup analysis for individual moderator levels. * This analysis did not include the adolescent population category, the general population category and the unknown population category as only two, one, and three effect sizes were available, respectively.
**Purpose:** Post hoc moderator/sensitivity results for measurement agreement. **Visual forms:** Complex table.
**Physical source pages:** 2, 44.
[PNG](<08_Robustness_and_Validation/P203-FigExtended%20Data%201-Extended%20Data%20Fig.%201%20Reporting%20accuracy%20post%20hoc%20moderator%20and%20subgroup%20analyses.%20Note%20k%20number%20of%20separate%20effect%20sizes%20included%20for%20the%20moderator....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=965>) · Original paper and literature note are not hosted.
### P203 - Fig. Extended Data 3
**Paper:** A Systematic Review and Meta-Analysis of Discrepancies Between Logged and Self-Reported Digital Media Use
**Original caption:** Extended Data Fig. 3 Digital media usage post hoc moderator and subgroup analyses. Note: k: number of separate effect sizes included for the moderator level; r = Pearson correlation coefficient; F values correspond to the Approximate Hotelling-Zhang with small sample correction omnibus tests for moderators with more than two levels; 95% CI corresponds to the r values for individual moderator levels; p corresponds to the F value for moderators or the subgroup analysis for individual moderator levels. * This analysis did not include the adolescent population group as only two effect sizes were available. † This analysis did not include the other category as only a single effect size was available.
**Purpose:** Post hoc moderator/sensitivity results for measurement agreement. **Visual forms:** Complex table.
**Physical source pages:** 2, 3, 43.
[PNG](<08_Robustness_and_Validation/P203-FigExtended%20Data%203-Extended%20Data%20Fig.%203%20Digital%20media%20usage%20post%20hoc%20moderator%20and%20subgroup%20analyses.%20Note%20k%20number%20of%20separate%20effect%20sizes%20included%20for%20the%20moderator....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=966>) · Original paper and literature note are not hosted.
### P204 - Fig. 2
**Paper:** Balancing Covariates via Propensity Score Weighting
**Original caption:** Figure 2: Boxplots for the absolute standardized differences for covariates under each weighting method in the RHC study.
**Purpose:** Covariate-balance diagnostic comparing weighting methods in the RHC application. **Visual forms:** Boxplot.
**Physical source pages:** 19.
[PNG](<08_Robustness_and_Validation/P204-Fig02-Boxplots%20for%20the%20absolute%20standardized%20differences%20for%20covariates%20under%20each%20weighting%20method%20in%20the%20RHC%20study.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=967>) · Original paper and literature note are not hosted.
### P204 - Fig. 3
**Paper:** Balancing Covariates via Propensity Score Weighting
**Original caption:** Figure 3: Distribution of the estimated propensity scores for White-Asian and White-Hispanic comparisons in the MEPS data.
**Purpose:** Propensity-score overlap diagnostic establishes limited support, not merely descriptive sample distribution. **Visual forms:** Density plot, Composite.
**Physical source pages:** 22.
[PNG](<08_Robustness_and_Validation/P204-Fig03-Distribution%20of%20the%20estimated%20propensity%20scores%20for%20White-Asian%20and%20White-Hispanic%20comparisons%20in%20the%20MEPS%20data.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=968>) · Original paper and literature note are not hosted.
### P204 - Fig. 4
**Paper:** Balancing Covariates via Propensity Score Weighting
**Original caption:** Figure 4: Unweighted and weighted BMI distributions for the White and Asian groups in the MEPS data.
**Purpose:** Distinct original four-panel weighted-covariate balance diagnostic was wrongly merged into Figure 3; separated while retaining its full caption. **Visual forms:** Density plot, Composite.
**Physical source pages:** 22.
[PNG](<08_Robustness_and_Validation/P204-Fig04-Unweighted%20and%20weighted%20BMI%20distributions%20for%20the%20White%20and%20Asian%20groups%20in%20the%20MEPS%20data.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=969>) · Original paper and literature note are not hosted.
### P204 - Fig. 5
**Paper:** Balancing Covariates via Propensity Score Weighting
**Original caption:** Figure 5: Boxplots for the absolute standardized difference for covariates under each weighting method.
**Purpose:** Covariate-balance diagnostics compare methods across two disparity applications. **Visual forms:** Boxplot, Composite.
**Physical source pages:** 23.
[PNG](<08_Robustness_and_Validation/P204-Fig05-Boxplots%20for%20the%20absolute%20standardized%20difference%20for%20covariates%20under%20each%20weighting%20method.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=970>) · Original paper and literature note are not hosted.
### P205 - Fig. 7
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** Figure 7 Essay scores of four groups before revision
**Purpose:** Baseline essay scores assess pre-intervention comparability; complete original caption restored. **Visual forms:** Boxplot, Scatter overlay.
**Physical source pages:** 41.
[PNG](<08_Robustness_and_Validation/P205-Fig07-Essay%20scores%20of%20four%20groups%20before%20revision.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=971>) · Original paper and literature note are not hosted.
### P211 - Table 5
**Paper:** Comparisons of statistical methods for handling attrition in a follow-up visit with complex survey sampling
**Original caption:** TABLE 5 Simulation results from difference models (linear regression) for various methods and missing mechanisms. Abbreviations: bias, average bias from 1000 simulations; CC, complete-case analysis; coverage, 95% confidence interval coverage rate from 1000 simulations; ESE, empirical standard deviation of the point estimates from 1000 simulations; FIML, full information maximum likelihood approach (full model); IPW, inverse probability weighting; MAR, missing-at-random; MCAR, missing-completely-at-random; MI, multiple imputation; MNAR, missing-not-at-random; NRCW, non-response cell weighting; SE, average standard error estimates from 1000 simulations.
**Purpose:** Exceptional multilevel simulation matrix partitions mechanisms and specification regimes, with method subcolumns under bias, standard error and coverage; continued pages form one table. **Visual forms:** Complex table.
**Physical source pages:** 14, 15.
[PNG](<08_Robustness_and_Validation/P211-Table05-Simulation%20results%20from%20difference%20models%20(linear%20regression)%20for%20various%20methods%20and%20missing%20mechanisms.%20Abbreviations%20bias,%20average%20bias%20from%201000%20simulations;....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=972>) · Original paper and literature note are not hosted.
### P211 - Table 6
**Paper:** Comparisons of statistical methods for handling attrition in a follow-up visit with complex survey sampling
**Original caption:** TABLE 6 Simulation results from rate of change models (linear regression) for various methods and missing mechanisms. Abbreviations: bias, average bias from 1000 simulations; CC, complete-case analysis; coverage, 95% confidence interval coverage rate from 1000 simulations; ESE, empirical standard deviation of the point estimates from 1000 simulations; FIML, full information maximum likelihood approach (full model); IPW, inverse probability weighting; MAR, missing-at-random; MCAR, missing-completely-at-random; MI, multiple imputation; MNAR, missing-not-at-random; NRCW, non-response cell weighting; SE, average standard error estimates from 1000 simulations.
**Purpose:** Exceptional multilevel simulation matrix partitions mechanisms and specification regimes, with method subcolumns under bias, standard error and coverage; continued pages form one table. **Visual forms:** Complex table.
**Physical source pages:** 16, 17.
[PNG](<08_Robustness_and_Validation/P211-Table06-Simulation%20results%20from%20rate%20of%20change%20models%20(linear%20regression)%20for%20various%20methods%20and%20missing%20mechanisms.%20Abbreviations%20bias,%20average%20bias%20from%201000....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=973>) · Original paper and literature note are not hosted.
### P211 - Table 7
**Paper:** Comparisons of statistical methods for handling attrition in a follow-up visit with complex survey sampling
**Original caption:** TABLE 7 Simulation results from binary models (logistic regression) for various methods and missing mechanisms. Abbreviations: bias, average bias from 1000 simulations; CC, complete-case analysis; coverage, 95% confidence interval coverage rate from 1000 simulations; ESE, empirical standard deviation of the point estimates from 1000 simulations; FIML, full information maximum likelihood approach (full model); IPW, inverse probability weighting; MAR, missing-at-random; MCAR, missing-completely-at-random; MI, multiple imputation; MNAR, missing-not-at-random; NRCW, non-response cell weighting; SE, average standard error estimates from 1000 simulations.
**Purpose:** Exceptional multilevel simulation matrix partitions mechanisms and specification regimes, with method subcolumns under bias, standard error and coverage; continued pages form one table. **Visual forms:** Complex table.
**Physical source pages:** 18, 19.
[PNG](<08_Robustness_and_Validation/P211-Table07-Simulation%20results%20from%20binary%20models%20(logistic%20regression)%20for%20various%20methods%20and%20missing%20mechanisms.%20Abbreviations%20bias,%20average%20bias%20from%201000%20simulations;%20CC,....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=974>) · Original paper and literature note are not hosted.
### P211 - Table 8
**Paper:** Comparisons of statistical methods for handling attrition in a follow-up visit with complex survey sampling
**Original caption:** TABLE 8 Simulation results from incidence models (Poisson regression) for various methods and missing mechanisms. Abbreviations: bias, average bias from 1000 simulations; CC, complete-case analysis; coverage, 95% confidence interval coverage rate from 1000 simulations; ESE, empirical standard deviation of the point estimates from 1000 simulations; FIML, full information maximum likelihood approach (full model); IPW, inverse probability weighting; MAR, missing-at-random; MCAR, missing-completely-at-random; MI, multiple imputation; MNAR, missing-not-at-random; NRCW, non-response cell weighting; SE, average standard error estimates from 1000 simulations.
**Purpose:** Exceptional multilevel simulation matrix crosses mechanism, model specification and coefficient rows against grouped method-performance columns. **Visual forms:** Complex table.
**Physical source pages:** 20.
[PNG](<08_Robustness_and_Validation/P211-Table08-Simulation%20results%20from%20incidence%20models%20(Poisson%20regression)%20for%20various%20methods%20and%20missing%20mechanisms.%20Abbreviations%20bias,%20average%20bias%20from%201000%20simulations;....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=975>) · Original paper and literature note are not hosted.
### P213 - Fig. 7
**Paper:** COVID-19 school closures, learning losses and intergenerational mobility
**Original caption:** Fig. 7 Relative mobility (correlation coefficient) under alternative loss index scenarios. Authors’ estimates based on data from Azevedo et al. (2022) and GDIM
**Purpose:** Sensitivity of projected relative mobility to alternative learning-loss-index assumptions. **Visual forms:** Dot plot.
**Physical source pages:** 9.
[PNG](<08_Robustness_and_Validation/P213-Fig07-Relative%20mobility%20(correlation%20coefficient)%20under%20alternative%20loss%20index%20scenarios.%20Authors%E2%80%99%20estimates%20based%20on%20data%20from%20Azevedo%20et%20al.%20(2022)%20and%20GDIM.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=976>) · Original paper and literature note are not hosted.
### P213 - App Fig. A2.2
**Paper:** COVID-19 school closures, learning losses and intergenerational mobility
**Original caption:** Figure A2.2: GDIM years of schooling and expected years of schooling (left panel) and learning-adjusted years of schooling (right panel). Source: Authors’ estimates based on data from Azevedo et al. (2022) and GDIM.
**Purpose:** Checks consistency of GDIM self-reported years with expected and learning-adjusted schooling measures. **Visual forms:** Scatter plot, Composite.
**Physical source pages:** 7.
[PNG](<08_Robustness_and_Validation/P213-App-FigA2.2-GDIM%20years%20of%20schooling%20and%20expected%20years%20of%20schooling%20(left%20panel)%20and%20learning-adjusted%20years%20of%20schooling%20(right%20panel).%20Source%20Authors%E2%80%99%20estimates%20based%20on....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=977>) · Original paper and literature note are not hosted.
### P215 - Fig. A1
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Appendix A1. Funnel plot of effect sizes for academic performance.
**Purpose:** Publication-bias/small-study-asymmetry diagnostic for this outcome; source appendix caption and all plot elements retained. **Visual forms:** Funnel plot.
**Physical source pages:** 23.
[PNG](<08_Robustness_and_Validation/P215-FigA1-Appendix%20A1.%20Funnel%20plot%20of%20effect%20sizes%20for%20academic%20performance.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=978>) · Original paper and literature note are not hosted.
### P215 - Fig. A2
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Appendix A2. Funnel plot of effect sizes for affective-motivational states.
**Purpose:** Publication-bias/small-study-asymmetry diagnostic for this outcome; source appendix caption and all plot elements retained. **Visual forms:** Funnel plot.
**Physical source pages:** 23.
[PNG](<08_Robustness_and_Validation/P215-FigA2-Appendix%20A2.%20Funnel%20plot%20of%20effect%20sizes%20for%20affective-motivational%20states.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=979>) · Original paper and literature note are not hosted.
### P215 - Fig. A3
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Appendix A3. Funnel plot of effect sizes for higher-order thinking propensities.
**Purpose:** Publication-bias/small-study-asymmetry diagnostic for this outcome; source appendix caption and all plot elements retained. **Visual forms:** Funnel plot.
**Physical source pages:** 24.
[PNG](<08_Robustness_and_Validation/P215-FigA3-Appendix%20A3.%20Funnel%20plot%20of%20effect%20sizes%20for%20higher-order%20thinking%20propensities.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=980>) · Original paper and literature note are not hosted.
### P215 - Fig. A4
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Appendix A4. Funnel plot of effect sizes for self-efficacy.
**Purpose:** Publication-bias/small-study-asymmetry diagnostic for this outcome; source appendix caption and all plot elements retained. **Visual forms:** Funnel plot.
**Physical source pages:** 24.
[PNG](<08_Robustness_and_Validation/P215-FigA4-Appendix%20A4.%20Funnel%20plot%20of%20effect%20sizes%20for%20self-efficacy.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=981>) · Original paper and literature note are not hosted.
### P215 - Fig. A5
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Appendix A5. Funnel plot of effect sizes for mental effort.
**Purpose:** Publication-bias/small-study-asymmetry diagnostic for this outcome; source appendix caption and all plot elements retained. **Visual forms:** Funnel plot.
**Physical source pages:** 25.
[PNG](<08_Robustness_and_Validation/P215-FigA5-Appendix%20A5.%20Funnel%20plot%20of%20effect%20sizes%20for%20mental%20effort.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=982>) · Original paper and literature note are not hosted.
### P222 - Fig. 4
**Paper:** From byproduct to design factor: on validating the interpretation of process indicators based on log data
**Original caption:** Fig. 4 Relationship between score group and proportion correct for literacy (selected item) (figure from Goldhammer et al., 2016, p. 24)
**Purpose:** Illustrates convergent validation of response-time engagement thresholds. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 17.
[PNG](<08_Robustness_and_Validation/P222-Fig04-Relationship%20between%20score%20group%20and%20proportion%20correct%20for%20literacy%20(selected%20item)%20(figure%20from%20Goldhammer%20et%20al.,%202016,%20p.%2024).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=983>) · Original paper and literature note are not hosted.
### P226 - Table 1
**Paper:** Granularity matters: comparing different ways of measuring self-regulated learning
**Original caption:** Table 1 Quality assessment of the studies included in the review
**Purpose:** Study-by-criterion quality appraisal matrix used to assess the review evidence. **Visual forms:** Complex table, Heatmap/matrix.
**Physical source pages:** 7.
[PNG](<08_Robustness_and_Validation/P226-Table01-Quality%20assessment%20of%20the%20studies%20included%20in%20the%20review.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=984>) · Original paper and literature note are not hosted.
### P227 - Fig. 2
**Paper:** How do test-takers rate their effort? A comparative analysis of self-report and log file data
**Original caption:** Fig. 2. Optimal cluster numbers for the elbow and silhouette methods.
**Purpose:** Selects the number of clusters using elbow and silhouette diagnostics. **Visual forms:** Line/profile plot, Composite.
**Physical source pages:** 28.
[PNG](<08_Robustness_and_Validation/P227-Fig02-Optimal%20cluster%20numbers%20for%20the%20elbow%20and%20silhouette%20methods.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=985>) · Original paper and literature note are not hosted.
### P230 - Fig. 4
**Paper:** Impact of AI assistance on student agency
**Original caption:** Fig. 4. Comparison of the control group (AI) and three experimental groups (NR, SR, and SAI) using six measures during the initial four weeks of AI prompt exposure. Boxplots represent data distribution with key statistics, while highlighted points indicate the mean. Violin plots provide an overview of data distribution using density curves.
**Purpose:** Initial common-AI phase verifies baseline behavioral comparability across later randomized groups; retain complete explanatory caption. **Visual forms:** Violin plot, Boxplot, Composite.
**Physical source pages:** 13.
[PNG](<08_Robustness_and_Validation/P230-Fig04-Comparison%20of%20the%20control%20group%20(AI)%20and%20three%20experimental%20groups%20(NR,%20SR,%20and%20SAI)%20using%20six%20measures%20during%20the%20initial%20four%20weeks%20of%20AI%20prompt%20exposure.%20Boxplots....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=986>) · Original paper and literature note are not hosted.
### P236 - Table B1
**Paper:** Measuring Test-Taking Effort on Constructed-Response Items with Item Response Time and Number of Actions
**Original caption:** Table B1. Informativeness (range of correlation coefficients between item scores with 10 plausible values) of response behaviours across effort measures-Stage I, first session.
**Purpose:** Color-highlighted item-by-method informativeness matrix identifies validation failures. **Visual forms:** Complex table.
**Physical source pages:** 21, 22.
[PNG](<08_Robustness_and_Validation/P236-TableB1-Informativeness%20(range%20of%20correlation%20coefficients%20between%20item%20scores%20with%2010%20plausible%20values)%20of%20response%20behaviours%20across%20effort%20measures-Stage%20I,%20first%20session.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=987>) · Original paper and literature note are not hosted.
### P236 - Table B2
**Paper:** Measuring Test-Taking Effort on Constructed-Response Items with Item Response Time and Number of Actions
**Original caption:** Table B2. Informativeness (range of correlation coefficients between item scores with 10 plausible values) of response behaviours across effort measures-Stage I, second session
**Purpose:** Color-highlighted item-by-method informativeness matrix identifies validation failures. **Visual forms:** Complex table.
**Physical source pages:** 23, 24.
[PNG](<08_Robustness_and_Validation/P236-TableB2-Informativeness%20(range%20of%20correlation%20coefficients%20between%20item%20scores%20with%2010%20plausible%20values)%20of%20response%20behaviours%20across%20effort%20measures-Stage%20I,%20second%20session.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=988>) · Original paper and literature note are not hosted.
### P236 - Fig. C1
**Paper:** Measuring Test-Taking Effort on Constructed-Response Items with Item Response Time and Number of Actions
**Original caption:** Graph C1. Boxplots of Number of Actions on Items from Stage I. Notes: NoA = number of actions. Items where NoA failed as an indicator of effort start with the letter “W”. The red dot stands for the mean NoA on an item.
**Purpose:** Post hoc checks whether action-count distributions explain invalid effort classifications. **Visual forms:** Density/histogram/boxplot.
**Physical source pages:** 25.
[PNG](<08_Robustness_and_Validation/P236-FigC1-Graph%20C1.%20Boxplots%20of%20Number%20of%20Actions%20on%20Items%20from%20Stage%20I.%20Notes%20NoA%20=%20number%20of%20actions.%20Items%20where%20NoA%20failed%20as%20an%20indicator%20of%20effort%20start%20with%20the%20letter....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=989>) · Original paper and literature note are not hosted.
### P236 - Fig. C2
**Paper:** Measuring Test-Taking Effort on Constructed-Response Items with Item Response Time and Number of Actions
**Original caption:** Graph C2. Examples of boxplots of Number of Actions by item score in Stage I. Note: NoA- Number of Actions
**Purpose:** Compares score-conditioned action distributions for items where the engagement indicator failed versus did not fail. **Visual forms:** Density/histogram/boxplot, Composite.
**Physical source pages:** 26.
[PNG](<08_Robustness_and_Validation/P236-FigC2-Graph%20C2.%20Examples%20of%20boxplots%20of%20Number%20of%20Actions%20by%20item%20score%20in%20Stage%20I.%20Note%20NoA-%20Number%20of%20Actions.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=990>) · Original paper and literature note are not hosted.
### P237 - Fig. 21.2
**Paper:** Modeling Item Responses with Response Times as Collateral Information in Large-Scale Educational Assessments
**Original caption:** Fig. 21.2 Scatter plots of estimated discrimination parameters under different models (left: 2 PM-pred vs. HM-pred, right: 2 PM-pred vs. eHM-pred)
**Purpose:** Checks how item discrimination changes under alternative measurement models. **Visual forms:** Scatter, Composite.
**Physical source pages:** 14.
[PNG](<08_Robustness_and_Validation/P237-Fig21.2-Scatter%20plots%20of%20estimated%20discrimination%20parameters%20under%20different%20models%20(left%202%20PM-pred%20vs.%20HM-pred,%20right%202%20PM-pred%20vs.%20eHM-pred).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=991>) · Original paper and literature note are not hosted.
### P238 - Fig. 4
**Paper:** Profiling low-proficiency science students in the Philippines using machine learning
**Original caption:** Fig. 4 Summary of accuracy and precision performance results for Random Forest model. The confusion matrix (a) and the ROC (b) summarizing the performance of the RF model in classifying the PISA 2018 Science Proficiency of Filipino students. The average accuracy is 0.74 and the area under the ROC being equal to 0.83.
**Purpose:** Held-out classification validation using confusion matrix and ROC. **Visual forms:** Heatmap/matrix, ROC curve, Composite.
**Physical source pages:** 6.
[PNG](<08_Robustness_and_Validation/P238-Fig04-Summary%20of%20accuracy%20and%20precision%20performance%20results%20for%20Random%20Forest%20model.%20The%20confusion%20matrix%20(a)%20and%20the%20ROC%20(b)%20summarizing%20the%20performance%20of%20the%20RF%20model....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=992>) · Original paper and literature note are not hosted.
### P238 - App Table U01
**Paper:** Profiling low-proficiency science students in the Philippines using machine learning
**Original caption:** Summary table of top 15 variables that have impact on each machine learning model. Note: Red font indicates positive impact on the prediction model, blue font indicates negative impact. Variables in bold font in the last four columns indicate common variable in the Random Forest model.
**Purpose:** Supplementary comparison of SHAP feature rankings across five algorithms; colour encodes positive/negative associations and bold encodes overlap with the preferred random forest. **Visual forms:** Colour-coded comparison, Complex table.
**Physical source pages:** 2.
[PNG](<08_Robustness_and_Validation/P238-App-TableU01-Summary%20table%20of%20top%2015%20variables%20that%20have%20impact%20on%20each%20machine%20learning%20model.%20Note%20Red%20font%20indicates%20positive%20impact%20on%20the%20prediction%20model,%20blue%20font....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=993>) · Original paper and literature note are not hosted.
### P243 - Fig. 2
**Paper:** Socioeconomic gradients in children’s cognitive skills: are cross-country comparisons robust to who reports family background?
**Original caption:** Figure 2. Estimated SES gradient in child test scores – comparison of results based on parent reports and child reports of father’s occupation (PISA) Notes: the graph shows the difference in average reading test scores, controlling for other variables included in the regression model, between children with the highest (‘high controller’) and lowest (‘semi/unskilled manual’) levels of father’s occupation. The units are national standard deviations. The correlation between the two sets of estimates is 0.94. Two letter country abbreviations are given in Table 4. The data refer to 2006. Final student weight applied.
**Purpose:** Sensitivity of cross-country SES gradients to choosing parent versus child reports. **Visual forms:** Scatter plot.
**Physical source pages:** 28.
[PNG](<08_Robustness_and_Validation/P243-Fig02-Estimated%20SES%20gradient%20in%20child%20test%20scores%20-%20comparison%20of%20results%20based%20on%20parent%20reports%20and%20child%20reports%20of%20father%E2%80%99s%20occupation%20(PISA)%20Notes%20the%20graph%20shows%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=994>) · Original paper and literature note are not hosted.
### P243 - Fig. 3
**Paper:** Socioeconomic gradients in children’s cognitive skills: are cross-country comparisons robust to who reports family background?
**Original caption:** Figure 3. Estimated SES gradient in child test scores – comparison of results based on parent reports and child reports of father’s education (PISA) Notes: the graph shows the difference in average reading test scores, controlling for other variables included in the regression model, between children with the highest (ISCED 5A+) and lowest (ISCED below 3A) levels of father’s education. The units are national standard deviations. The correlation between the two sets of estimates is 0.89. Two letter country abbreviations are given in Table 3; the digit ending the label shows the year to which the data point refers (‘6’ = 2006, ‘9’ = 2009). Final student weights applied.
**Purpose:** Sensitivity of cross-country SES gradients to choosing parent versus child reports. **Visual forms:** Scatter plot.
**Physical source pages:** 29.
[PNG](<08_Robustness_and_Validation/P243-Fig03-Estimated%20SES%20gradient%20in%20child%20test%20scores%20-%20comparison%20of%20results%20based%20on%20parent%20reports%20and%20child%20reports%20of%20father%E2%80%99s%20education%20(PISA)%20Notes%20the%20graph%20shows%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=995>) · Original paper and literature note are not hosted.
### P243 - Fig. 4
**Paper:** Socioeconomic gradients in children’s cognitive skills: are cross-country comparisons robust to who reports family background?
**Original caption:** Figure 4. Estimated SES gradient in child test scores – comparison of results based on parent reports and child reports of books in the home (PIRLS). Notes: the graph shows the difference in average reading test scores, controlling for other variables included in the regression model, between children with the highest (more than 200 books) and lowest (0-25 books) reported levels of books in the home. The units are national standard deviations. The correlation between the two sets of estimates is 0.37. Two letter country abbreviations are given in Table 3; the digit ending the label shows the year to which the data point refers (‘1’ = 2001, ‘6’ = 2006). Final student weights applied.
**Purpose:** Sensitivity of cross-country SES gradients to choosing parent versus child reports. **Visual forms:** Scatter plot.
**Physical source pages:** 30.
[PNG](<08_Robustness_and_Validation/P243-Fig04-Estimated%20SES%20gradient%20in%20child%20test%20scores%20-%20comparison%20of%20results%20based%20on%20parent%20reports%20and%20child%20reports%20of%20books%20in%20the%20home%20(PIRLS).%20Notes%20the%20graph%20shows%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=996>) · Original paper and literature note are not hosted.
### P243 - Table 5
**Paper:** Socioeconomic gradients in children’s cognitive skills: are cross-country comparisons robust to who reports family background?
**Original caption:** Table 5. Correlations between country estimates of SES gradients in reading test scores based on parent and child reports: the impact of changing the comparison group. Notes: SES gradients in reading test score have been estimated using every possible pair of categories within each of the measures. The categories in the first column are the changing comparison categories. For each SES measure, the figures give the correlations between the two sets of estimates of gradients, one based on child reports of SES and one on parent reports. For example, the circled figure in panel (a) refers to the gradients estimated using the ‘high professional’ and ‘semi/unskilled manual’ groups (that is, the difference in average reading test scores between these two categories, controlling for other variables in the regression model). The circled figures in each panel give the correlations for the scattergrams in Figures 2 and 4. Final student weights applied. Results for occupation and education refer to PISA. Results for books in the home refer to PIRLS.
**Purpose:** Recovered by direct visual comparison with source page. **Visual forms:** Matrix, Composite, Complex table.
**Physical source pages:** 26.
[PNG](<08_Robustness_and_Validation/P243-Table05-Correlations%20between%20country%20estimates%20of%20SES%20gradients%20in%20reading%20test%20scores%20based%20on%20parent%20and%20child%20reports%20the%20impact%20of%20changing%20the%20comparison%20group.%20Notes....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=997>) · Original paper and literature note are not hosted.
### P253 - App Fig. S2
**Paper:** Validity and fairness of the PISA 2018 Global Competence assessment: an argument-based evaluation via explanatory item response models
**Original caption:** Figure S2. Percentage differences in variance components between 2-level and 3-level models by booklet group
**Purpose:** Sensitivity of estimated variance components to using two-level versus three-level models. **Visual forms:** Grouped bar chart.
**Physical source pages:** 7.
[PNG](<08_Robustness_and_Validation/P253-App-FigS2-Percentage%20differences%20in%20variance%20components%20between%202-level%20and%203-level%20models%20by%20booklet%20group.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=998>) · Original paper and literature note are not hosted.
### P253 - App Fig. S4
**Paper:** Validity and fairness of the PISA 2018 Global Competence assessment: an argument-based evaluation via explanatory item response models
**Original caption:** Figure S4. Funnel plot for self-efficacy (Level-2 predictor) Note. Given the limited number of booklet groups (k = 10), formal tests of funnel plot asymmetry (e.g., Egger regression) were not performed in accordance with methodological recommendations. Visual inspection was considered sufficient for diagnostic purposes.
**Purpose:** Exploratory asymmetry/heterogeneity diagnostic across ten booklet groups; original shared methodological note retained. **Visual forms:** Funnel plot.
**Physical source pages:** 19, 20.
[PNG](<08_Robustness_and_Validation/P253-App-FigS4-Funnel%20plot%20for%20self-efficacy%20(Level-2%20predictor)%20Note.%20Given%20the%20limited%20number%20of%20booklet%20groups%20(k%20=%2010),%20formal%20tests%20of%20funnel%20plot%20asymmetry%20(e.g.,%20Egger....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=999>) · Original paper and literature note are not hosted.
### P253 - App Fig. S5
**Paper:** Validity and fairness of the PISA 2018 Global Competence assessment: an argument-based evaluation via explanatory item response models
**Original caption:** Figure S5. Funnel plot for respect for people from other cultural backgrounds (Level-2 predictor) Note. Given the limited number of booklet groups (k = 10), formal tests of funnel plot asymmetry (e.g., Egger regression) were not performed in accordance with methodological recommendations. Visual inspection was considered sufficient for diagnostic purposes.
**Purpose:** Exploratory asymmetry/heterogeneity diagnostic across ten booklet groups; original shared methodological note retained. **Visual forms:** Funnel plot.
**Physical source pages:** 20.
[PNG](<08_Robustness_and_Validation/P253-App-FigS5-Funnel%20plot%20for%20respect%20for%20people%20from%20other%20cultural%20backgrounds%20(Level-2%20predictor)%20Note.%20Given%20the%20limited%20number%20of%20booklet%20groups%20(k%20=%2010),%20formal%20tests%20of....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1000>) · Original paper and literature note are not hosted.
### P253 - App Fig. S6
**Paper:** Validity and fairness of the PISA 2018 Global Competence assessment: an argument-based evaluation via explanatory item response models
**Original caption:** Figure S6. Funnel plot for attitudes toward immigrants (Level-2 predictor). Note. Given the limited number of booklet groups (k = 10), formal tests of funnel plot asymmetry (e.g., Egger regression) were not performed in accordance with methodological recommendations. Visual inspection was considered sufficient for diagnostic purposes.
**Purpose:** Exploratory asymmetry/heterogeneity diagnostic across ten booklet groups; original shared methodological note retained. **Visual forms:** Funnel plot.
**Physical source pages:** 20.
[PNG](<08_Robustness_and_Validation/P253-App-FigS6-Funnel%20plot%20for%20attitudes%20toward%20immigrants%20(Level-2%20predictor).%20Note.%20Given%20the%20limited%20number%20of%20booklet%20groups%20(k%20=%2010),%20formal%20tests%20of%20funnel%20plot%20asymmetry....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1001>) · Original paper and literature note are not hosted.
### P253 - App Fig. S9
**Paper:** Validity and fairness of the PISA 2018 Global Competence assessment: an argument-based evaluation via explanatory item response models
**Original caption:** Figure S9. Funnel plot grid for items exhibiting moderate and large gender-related DIF
**Purpose:** Exploratory consistency diagnostics for items with moderate or large gender DIF. **Visual forms:** Funnel plot, Composite.
**Physical source pages:** 25.
[PNG](<08_Robustness_and_Validation/P253-App-FigS9-Funnel%20plot%20grid%20for%20items%20exhibiting%20moderate%20and%20large%20gender-related%20DIF.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1002>) · Original paper and literature note are not hosted.
## 09 Literature Mapping
### P082 - Fig. 2
**Paper:** National contexts and information and communication technology in education: A systematic review of PISA studies
**Original caption:** Fig. 2. Number of Published Articles by the Year and Cycles of Data This figure displays the number of articles published each year and aligns them with corresponding PISA assessment cycles (2000–2018).
**Purpose:** Publication-year and PISA-cycle map of the reviewed literature. **Visual forms:** Bar chart.
**Physical source pages:** 5.
[PNG](<09_Literature_Mapping/P082-Fig02-Number%20of%20Published%20Articles%20by%20the%20Year%20and%20Cycles%20of%20Data%20This%20figure%20displays%20the%20number%20of%20articles%20published%20each%20year%20and%20aligns%20them%20with%20corresponding%20PISA....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=690>) · Original paper and literature note are not hosted.
### P082 - Fig. 3
**Paper:** National contexts and information and communication technology in education: A systematic review of PISA studies
**Original caption:** Fig. 3. Number of Articles by Type of Questionnaire This figure shows the distribution of included studies based on the type of questionnaire used.
**Purpose:** Maps questionnaire data sources used by reviewed studies. **Visual forms:** Bar chart.
**Physical source pages:** 6.
[PNG](<09_Literature_Mapping/P082-Fig03-Number%20of%20Articles%20by%20Type%20of%20Questionnaire%20This%20figure%20shows%20the%20distribution%20of%20included%20studies%20based%20on%20the%20type%20of%20questionnaire%20used.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=691>) · Original paper and literature note are not hosted.
### P082 - Fig. 4
**Paper:** National contexts and information and communication technology in education: A systematic review of PISA studies
**Original caption:** Fig. 4. Distribution of Conceptual Overarching Themes This figure presents the categorization of articles into major conceptual themes, such as access, use, competence, or country focus.
**Purpose:** Maps conceptual themes in the reviewed research corpus. **Visual forms:** Bar chart.
**Physical source pages:** 7.
[PNG](<09_Literature_Mapping/P082-Fig04-Distribution%20of%20Conceptual%20Overarching%20Themes%20This%20figure%20presents%20the%20categorization%20of%20articles%20into%20major%20conceptual%20themes,%20such%20as%20access,%20use,%20competence,%20or%20country....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=692>) · Original paper and literature note are not hosted.
### P082 - App Fig. C1
**Paper:** C017
**Original caption:** Figure C1. Country Affiliation of the First Author
**Purpose:** Geographic authorship distribution of the reviewed literature; caption taken from original figure heading, not the earlier table of contents. **Visual forms:** Bar chart.
**Physical source pages:** 14.
[PNG](<09_Literature_Mapping/P082-App-FigC1-Country%20Affiliation%20of%20the%20First%20Author.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1003>) · Original paper and literature note are not hosted.
### P082 - App Table C2
**Paper:** C017
**Original caption:** Table C2. Country-level Factors Investigated
**Purpose:** Categorized evidence inventory links country-level constructs to their original data sources; retain all thematic groupings and continuation. **Visual forms:** Complex table.
**Physical source pages:** 9, 10.
[PNG](<09_Literature_Mapping/P082-App-TableC2-Country-level%20Factors%20Investigated.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1004>) · Original paper and literature note are not hosted.
### P082 - App Table C3
**Paper:** C017
**Original caption:** Table C3. Overview of Learning Outcomes, ICC Values, and Predictors. Note. This table summarizes 17 studies that reported intraclass correlation coefficients (ICCs) among the 29 studies included in the review. Only studies that employed multilevel analyses and provided ICC values for learning outcomes were included. Student-level ICT-related factors are grouped into four categories (ICT access, ICT use, ICT attitudes, and others) rather than listing individual variables.
**Purpose:** Multilevel evidence matrix maps each study and outcome to country-level and student-level predictors; all three pages and scope note are one original table. **Visual forms:** Complex table.
**Physical source pages:** 11, 12, 13.
[PNG](<09_Literature_Mapping/P082-App-TableC3-Overview%20of%20Learning%20Outcomes,%20ICC%20Values,%20and%20Predictors.%20Note.%20This%20table%20summarizes%2017%20studies%20that%20reported%20intraclass%20correlation%20coefficients%20(ICCs)%20among%20the....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1005>) · Original paper and literature note are not hosted.
### P129 - Fig. 3
**Paper:** A bibliometric and thematic analysis of PISA literature for school improvement and effectiveness over two decades
**Original caption:** Fig. 3. Longitudinal evolution of PISA literature, 2001–2024 (N = 3729)
**Purpose:** Annual publication growth in PISA scholarship. **Visual forms:** Line/profile plot.
**Physical source pages:** 5.
[PNG](<09_Literature_Mapping/P129-Fig03-Longitudinal%20evolution%20of%20PISA%20literature,%202001-2024%20(N%20=%203729).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=693>) · Original paper and literature note are not hosted.
### P129 - Fig. 4
**Paper:** A bibliometric and thematic analysis of PISA literature for school improvement and effectiveness over two decades
**Original caption:** Fig. 4. Worldwide distribution of PISA literature, 2001–2024 (N = 3729)
**Purpose:** Geographic distribution of PISA literature. **Visual forms:** Map.
**Physical source pages:** 6.
[PNG](<09_Literature_Mapping/P129-Fig04-Worldwide%20distribution%20of%20PISA%20literature,%202001-2024%20(N%20=%203729).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=694>) · Original paper and literature note are not hosted.
### P129 - Fig. 5
**Paper:** A bibliometric and thematic analysis of PISA literature for school improvement and effectiveness over two decades
**Original caption:** Fig. 5. Intellectual structure of PISA literature, 2001–2024 (N = 3729).
**Purpose:** Intellectual/co-citation structure of PISA scholarship. **Visual forms:** Network.
**Physical source pages:** 7.
[PNG](<09_Literature_Mapping/P129-Fig05-Intellectual%20structure%20of%20PISA%20literature,%202001-2024%20(N%20=%203729).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=695>) · Original paper and literature note are not hosted.
### P129 - Fig. 6
**Paper:** A bibliometric and thematic analysis of PISA literature for school improvement and effectiveness over two decades
**Original caption:** Fig. 6. Conceptual themes of PISA literature, 2001–2024 (N = 3729).
**Purpose:** Keyword themes in PISA literature. **Visual forms:** Network.
**Physical source pages:** 8.
[PNG](<09_Literature_Mapping/P129-Fig06-Conceptual%20themes%20of%20PISA%20literature,%202001-2024%20(N%20=%203729).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=696>) · Original paper and literature note are not hosted.
### P129 - Fig. 7
**Paper:** A bibliometric and thematic analysis of PISA literature for school improvement and effectiveness over two decades
**Original caption:** Fig. 7. Development patterns for the conceptual themes of PISA literature, 2001–2024 (N = 3729).
**Purpose:** Temporal development of mapped literature themes. **Visual forms:** Flowchart, Complex table.
**Physical source pages:** 9.
[PNG](<09_Literature_Mapping/P129-Fig07-Development%20patterns%20for%20the%20conceptual%20themes%20of%20PISA%20literature,%202001-2024%20(N%20=%203729).png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=697>) · Original paper and literature note are not hosted.
### P129 - Fig. 8
**Paper:** A bibliometric and thematic analysis of PISA literature for school improvement and effectiveness over two decades
**Original caption:** Fig. 8. Research front of PISA literature, 2019–2024.
**Purpose:** Recent research fronts in PISA literature. **Visual forms:** Network.
**Physical source pages:** 10.
[PNG](<09_Literature_Mapping/P129-Fig08-Research%20front%20of%20PISA%20literature,%202019-2024.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=698>) · Original paper and literature note are not hosted.
### P135 - Fig. 1
**Paper:** The application of machine learning approaches in international large-scale educational assessments: A scoping review
**Original caption:** Fig. 1. Systematic search procedure.
**Purpose:** Bar chart maps frequencies of machine-learning approaches in reviewed studies; source caption mismatch retained. **Visual forms:** Bar chart.
**Physical source pages:** 5.
[PNG](<09_Literature_Mapping/P135-Fig01-Systematic%20search%20procedure.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=699>) · Original paper and literature note are not hosted.
### P135 - Table 2
**Paper:** The application of machine learning approaches in international large-scale educational assessments: A scoping review
**Original caption:** Table 2 Theoretical frameworks adopted in the included articles. Note. Although the PPCT model stems from socio-ecological theory, we list it separately due to their nuanced differences. Unlike traditional socio-ecological frameworks, the PPCT model highlights the dynamic interplay of process, person, context, and time in human development, integrating individual agency and the evolving nature of development.
**Purpose:** Structured evidence map links theoretical frameworks and their adopting studies. **Visual forms:** Complex table.
**Physical source pages:** 7.
[PNG](<09_Literature_Mapping/P135-Table02-Theoretical%20frameworks%20adopted%20in%20the%20included%20articles.%20Note.%20Although%20the%20PPCT%20model%20stems%20from%20socio-ecological%20theory,%20we%20list%20it%20separately%20due%20to%20their%20nuanced....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=700>) · Original paper and literature note are not hosted.
### P188 - Table 1
**Paper:** Socioeconomic status and student learning: insights from an umbrella review
**Original caption:** Table 1 Focus of reviews
**Purpose:** Evidence matrix maps included reviews across focus, populations, outcomes, SES indicators and mechanisms. **Visual forms:** Complex table.
**Physical source pages:** 4, 5, 6, 7, 8, 9, 10.
[PNG part 1](<09_Literature_Mapping/P188-Table01-Focus%20of%20reviews-Part01.png>) · [PNG part 2](<09_Literature_Mapping/P188-Table01-Focus%20of%20reviews-Part02.png>) · [PNG part 3](<09_Literature_Mapping/P188-Table01-Focus%20of%20reviews-Part03.png>) · [PNG part 4](<09_Literature_Mapping/P188-Table01-Focus%20of%20reviews-Part04.png>) · [PNG part 5](<09_Literature_Mapping/P188-Table01-Focus%20of%20reviews-Part05.png>) · [PNG part 6](<09_Literature_Mapping/P188-Table01-Focus%20of%20reviews-Part06.png>) · [PNG part 7](<09_Literature_Mapping/P188-Table01-Focus%20of%20reviews-Part07.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=701>) · Original paper and literature note are not hosted.
### P215 - Table 1
**Paper:** Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies
**Original caption:** Table 1. Previous reviews on ChatGPT in education. Note: If publication date is not provided, the acceptable date is recorded.
**Purpose:** Structured evidence comparison across review publication dates, databases, search windows, study counts and contributions. **Visual forms:** Complex table.
**Physical source pages:** 4.
[PNG](<09_Literature_Mapping/P215-Table01-Previous%20reviews%20on%20ChatGPT%20in%20education.%20Note%20If%20publication%20date%20is%20not%20provided,%20the%20acceptable%20date%20is%20recorded.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1006>) · Original paper and literature note are not hosted.
## 10 Experimental Materials
### P065 - Fig. 3
**Paper:** The impact of CT-integrated mathematics instruction on secondary school students' computational thinking skills
**Original caption:** Fig. 3. Conceptual learning and problem-solving performance of the experimental class.
**Purpose:** Photographs document the classroom activities used as intervention materials. **Visual forms:** Photograph, Composite.
**Physical source pages:** 5.
[PNG](<10_Experimental_Materials/P065-Fig03-Conceptual%20learning%20and%20problem-solving%20performance%20of%20the%20experimental%20class.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=708>) · Original paper and literature note are not hosted.
### P123 - Table 2
**Paper:** Professional Development of New Higher Education Teachers With Information and Communication Technology in Shanghai: A Kirkpatrick’s Evaluation Approach
**Original caption:** Table 2. Description of Teaching with ICT Training Module.
**Purpose:** Structured schedule and activity matrix describing the ICT training intervention. **Visual forms:** Complex table.
**Physical source pages:** 11.
[PNG](<10_Experimental_Materials/P123-Table02-Description%20of%20Teaching%20with%20ICT%20Training%20Module.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=709>) · Original paper and literature note are not hosted.
### P134 - Fig. 2
**Paper:** Separating prior knowledge from acquired knowledge: An individual differences analysis of PISA - learning in the digital world
**Original caption:** Fig. 2. Karel the Turtle – Example Task of a “Block-based Programming“Unit. Note. Students build a program to control the turtle (1), can check the functionality of their block script (2), and use learning materials to further improve their program (3).
**Purpose:** Illustrated block-programming task provides experimental material. **Visual forms:** Interface.
**Physical source pages:** 4.
[PNG](<10_Experimental_Materials/P134-Fig02-Karel%20the%20Turtle%20-%20Example%20Task%20of%20a%20%E2%80%9CBlock-based%20Programming%E2%80%9CUnit.%20Note.%20Students%20build%20a%20program%20to%20control%20the%20turtle%20(1),%20can%20check%20the%20functionality%20of%20their%20block....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=710>) · Original paper and literature note are not hosted.
### P134 - Fig. 3
**Paper:** Separating prior knowledge from acquired knowledge: An individual differences analysis of PISA - learning in the digital world
**Original caption:** Fig. 3. Movie Ratings – Example Task of a “System Modelling“Unit. Note. Students manipulate the ticket price and the distance to cinema to observe the effect on movie ratings (1), select the discovered relationships (2) and use learning materials to further improve their solution (3).
**Purpose:** Illustrated system-modeling task and example support provide experimental material. **Visual forms:** Interface.
**Physical source pages:** 5.
[PNG](<10_Experimental_Materials/P134-Fig03-Movie%20Ratings%20-%20Example%20Task%20of%20a%20%E2%80%9CSystem%20Modelling%E2%80%9CUnit.%20Note.%20Students%20manipulate%20the%20ticket%20price%20and%20the%20distance%20to%20cinema%20to%20observe%20the%20effect%20on%20movie%20ratings....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=711>) · Original paper and literature note are not hosted.
### P156 - Table 1
**Paper:** Effects of LLM use and note-taking on reading comprehension and memory: A randomised experiment in secondary schools
**Original caption:** Table 1. Question types and scoring for literal retention, comprehension, and free recall. Note. Two of the eight “Short response - Cued recall” questions for literal retention are worth two points each.
**Purpose:** Merged outcome partitions compare experimental test questions, scoring rubrics, and maximum scores. **Visual forms:** Complex table.
**Physical source pages:** 8.
[PNG](<10_Experimental_Materials/P156-Table01-Question%20types%20and%20scoring%20for%20literal%20retention,%20comprehension,%20and%20free%20recall.%20Note.%20Two%20of%20the%20eight%20%E2%80%9CShort%20response%20-%20Cued%20recall%E2%80%9D%20questions%20for%20literal....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=712>) · Original paper and literature note are not hosted.
### P156 - Fig. 2
**Paper:** Effects of LLM use and note-taking on reading comprehension and memory: A randomised experiment in secondary schools
**Original caption:** Fig. 2. Example task screen for the LLM + Notes condition.
**Purpose:** Actual reading, chatbot, and notepad experimental interface. **Visual forms:** Interface.
**Physical source pages:** 8.
[PNG](<10_Experimental_Materials/P156-Fig02-Example%20task%20screen%20for%20the%20LLM%20+%20Notes%20condition.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=713>) · Original paper and literature note are not hosted.
### P159 - Fig. A2
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure A2: Treatment Instructions Displayed to Participants. Notes: This figure shows the treatment-specific instructions displayed to participants in the survey interface.
**Purpose:** Shows the treatment-specific participant instructions in the survey interface. **Visual forms:** Composite.
**Physical source pages:** 44.
[PNG](<10_Experimental_Materials/P159-FigA2-Treatment%20Instructions%20Displayed%20to%20Participants.%20Notes%20This%20figure%20shows%20the%20treatment-specific%20instructions%20displayed%20to%20participants%20in%20the%20survey%20interface.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=714>) · Original paper and literature note are not hosted.
### P159 - Fig. B1
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure B1: Automation Example: Student Delegates Wr Notes: This figure shows an excerpt of a ChatGPT conversation classified as automation. The student’s first prompt is verbatim. The bracketed text summarizes the other turns.
**Purpose:** Shows a participant interaction example used to illustrate automation in the experiment. **Visual forms:** Interface, Composite.
**Physical source pages:** 69.
[PNG](<10_Experimental_Materials/P159-FigB1-Automation%20Example%20Student%20Delegates%20Wr%20Notes%20This%20figure%20shows%20an%20excerpt%20of%20a%20ChatGPT%20conversation%20classified%20as%20automation.%20The%20student%E2%80%99s%20first%20prompt%20is%20verbatim....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=715>) · Original paper and literature note are not hosted.
### P159 - Fig. B2
**Paper:** Experimental Evidence on the Learning Impact of Generative AI
**Original caption:** Figure B2: Augmentation Example: Student Uses ChatGPT as an Explainer.
**Purpose:** Shows an example of the AI-assisted learning interaction used to illustrate augmentation. **Visual forms:** Interface, Composite.
**Physical source pages:** 70.
[PNG](<10_Experimental_Materials/P159-FigB2-Augmentation%20Example%20Student%20Uses%20ChatGPT%20as%20an%20Explainer.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=716>) · Original paper and literature note are not hosted.
### P195 - Fig. U01
**Paper:** Trends in Income Nonresponse Over Two Decades
**Original caption:** No original caption supplied.
**Purpose:** Original unfolding-income-bracket questionnaire routing material. **Visual forms:** Flowchart.
**Physical source pages:** 17.
[PNG](<10_Experimental_Materials/P195-FigU01-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=717>) · Original paper and literature note are not hosted.
### P205 - Fig. 2
**Paper:** Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance
**Original caption:** FIGURE 2 Learning environment and learning tools
**Purpose:** Complete learning interface plus three support-tool examples; caption visually transcribed from source image. **Visual forms:** Interface, Composite.
**Physical source pages:** 9.
[PNG](<10_Experimental_Materials/P205-Fig02-Learning%20environment%20and%20learning%20tools.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1007>) · Original paper and literature note are not hosted.
### P210 - Fig. 1
**Paper:** Cognitive Offloading
**Original caption:** Figure 1 Caption. Paradigms for investigating cognitive offloading. Panel A: In the external normalization paradigm [9] participants read arrays of words that are presented in upright or rotated orientations. When faced with rotated words, participants can align them using internal cognitive processes (‘internal normalization’) or physical action (‘external normalization’). Panel B: In the intention offloading paradigm [10] participants use a mouse or touchscreen to drag numbered circles in sequence to the bottom of the screen. They are also instructed at the beginning of the trial that one or more of these circles should be dragged to an alternative location. They can either remember these intentions internally or offload them by dragging target circles towards their intended location at the beginning of the trial. In some ways this is analogous to everyday offloading behavior such as leaving an item by the front door so that we will remember it when leaving the house. For a demonstration of the task, please visit “http://samgilbert.net/offloadDemo.html”. Illustrations modified with permission from [9] and [47].
**Purpose:** Both external-normalization and intention-offloading panels form the same original figure; detached full caption retained as source pieces. **Visual forms:** Composite, Experimental paradigm.
**Physical source pages:** 26, 27, 35.
[PNG](<10_Experimental_Materials/P210-Fig01-Caption.%20Paradigms%20for%20investigating%20cognitive%20offloading.%20Panel%20A%20In%20the%20external%20normalization%20paradigm%20%5B9%5D%20participants%20read%20arrays%20of%20words%20that%20are%20presented%20in....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1008>) · Original paper and literature note are not hosted.
### P219 - Fig. 2
**Paper:** Explaining social disparities in ICT literacy: the role of multiple socioeconomic status indicators and ICT-related motivations
**Original caption:** Fig. 2 Example item of the ICT literacy test. Copyright Leibniz Institute for Educational Trajectories (LIfBi). Reproduced with permission
**Purpose:** Complete original ICT test example: instructions, spreadsheet stimulus, all response options, copyright and caption. **Visual forms:** Example test item, Interface.
**Physical source pages:** 12.
[PNG](<10_Experimental_Materials/P219-Fig02-Example%20item%20of%20the%20ICT%20literacy%20test.%20Copyright%20Leibniz%20Institute%20for%20Educational%20Trajectories%20(LIfBi).%20Reproduced%20with%20permission.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1009>) · Original paper and literature note are not hosted.
### P222 - Fig. 1
**Paper:** From byproduct to design factor: on validating the interpretation of process indicators based on log data
**Original caption:** Fig. 1 PIAAC Job Search unit (Item 1). The screenshot shows the search engine results page presented to the test-taker at the beginning. To solve the item task, test-takers can navigate to the linked pages
**Purpose:** Caption/context matches screenshot|interface|experimental material|example item|example.*task|stimulus|learning material. **Visual forms:** Interface.
**Physical source pages:** 5.
[PNG](<10_Experimental_Materials/P222-Fig01-PIAAC%20Job%20Search%20unit%20(Item%201).%20The%20screenshot%20shows%20the%20search%20engine%20results%20page%20presented%20to%20the%20test-taker%20at%20the%20beginning.%20To%20solve%20the%20item%20task,%20test-takers....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1010>) · Original paper and literature note are not hosted.
### P222 - Fig. 5
**Paper:** From byproduct to design factor: on validating the interpretation of process indicators based on log data
**Original caption:** Fig. 5 Example unit assessing multiple document comprehension. The screenshot shows the dialog box with the source information after the reader has clicked on the source button (3)
**Purpose:** Caption/context matches screenshot|interface|experimental material|example item|example.*task|stimulus|learning material. **Visual forms:** Interface.
**Physical source pages:** 18.
[PNG](<10_Experimental_Materials/P222-Fig05-Example%20unit%20assessing%20multiple%20document%20comprehension.%20The%20screenshot%20shows%20the%20dialog%20box%20with%20the%20source%20information%20after%20the%20reader%20has%20clicked%20on%20the%20source....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1011>) · Original paper and literature note are not hosted.
### P224 - Fig. 1
**Paper:** Generative AI without guardrails can harm learning: Evidence from high school mathematics
**Original caption:** Fig. 1. Prompts used in GPT Base and GPT Tutor for the first 11th grade practice problem in the first session.
**Purpose:** Original numbered experimental prompt comparison is a scientific figure despite being text-heavy; full two-prompt content retained. **Visual forms:** Text-based experimental material.
**Physical source pages:** 3.
[PNG](<10_Experimental_Materials/P224-Fig01-Prompts%20used%20in%20GPT%20Base%20and%20GPT%20Tutor%20for%20the%20first%2011th%20grade%20practice%20problem%20in%20the%20first%20session.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1012>) · Original paper and literature note are not hosted.
### P227 - Fig. 1
**Paper:** How do test-takers rate their effort? A comparative analysis of self-report and log file data
**Original caption:** Fig. 1. The first stage of interactive problem-solving: exploring the relationship between two input and two output variables.
**Purpose:** Illustrates the MicroDYN task interface. **Visual forms:** Interface.
**Physical source pages:** 27.
[PNG](<10_Experimental_Materials/P227-Fig01-The%20first%20stage%20of%20interactive%20problem-solving%20exploring%20the%20relationship%20between%20two%20input%20and%20two%20output%20variables.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1013>) · Original paper and literature note are not hosted.
### P228 - Fig. 3
**Paper:** How to conceptualize, represent, and analyze log data from technology-based assessments? A generic framework and an application to questionnaire items
**Original caption:** Fig. 3 Example CBA screen of the PISA 2015 context questionnaire (schematic view) for an item battery with 3 items
**Purpose:** Illustrates assessment interface. **Visual forms:** Interface.
**Physical source pages:** 14.
[PNG](<10_Experimental_Materials/P228-Fig03-Example%20CBA%20screen%20of%20the%20PISA%202015%20context%20questionnaire%20(schematic%20view)%20for%20an%20item%20battery%20with%203%20items.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1014>) · Original paper and literature note are not hosted.
### P230 - Fig. 1
**Paper:** Impact of AI assistance on student agency
**Original caption:** Fig. 1. Interfaces of peer review processes in RiPPLE.
**Purpose:** Original peer-review interface and feedback-outcome display used by learners. **Visual forms:** Interface, Composite.
**Physical source pages:** 4.
[PNG](<10_Experimental_Materials/P230-Fig01-Interfaces%20of%20peer%20review%20processes%20in%20RiPPLE.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1015>) · Original paper and literature note are not hosted.
### P230 - Fig. 2
**Paper:** Impact of AI assistance on student agency
**Original caption:** Fig. 2. Complementary peer review interfaces include (a) AI-assistance prompts and (b) a self-monitoring checklist.
**Purpose:** Retains AI-prompt examples, checklist interface, panel labels and caption as one original figure. **Visual forms:** Interface, Composite.
**Physical source pages:** 5.
[PNG](<10_Experimental_Materials/P230-Fig02-Complementary%20peer%20review%20interfaces%20include%20(a)%20AI-assistance%20prompts%20and%20(b)%20a%20self-monitoring%20checklist.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1016>) · Original paper and literature note are not hosted.
### P235 - App Fig. U01
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 1. **Visual forms:** Interface.
**Physical source pages:** 42.
[PNG](<10_Experimental_Materials/P235-AppB-FigU01-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1017>) · Original paper and literature note are not hosted.
### P235 - App Fig. U02
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 2. **Visual forms:** Interface.
**Physical source pages:** 43.
[PNG](<10_Experimental_Materials/P235-AppB-FigU02-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1018>) · Original paper and literature note are not hosted.
### P235 - App Fig. U03
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 3. **Visual forms:** Interface.
**Physical source pages:** 44.
[PNG](<10_Experimental_Materials/P235-AppB-FigU03-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1019>) · Original paper and literature note are not hosted.
### P235 - App Fig. U04
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 4. **Visual forms:** Interface.
**Physical source pages:** 45.
[PNG](<10_Experimental_Materials/P235-AppB-FigU04-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1020>) · Original paper and literature note are not hosted.
### P235 - App Fig. U05
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 5. **Visual forms:** Interface.
**Physical source pages:** 46.
[PNG](<10_Experimental_Materials/P235-AppB-FigU05-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1021>) · Original paper and literature note are not hosted.
### P235 - App Fig. U06
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 6. **Visual forms:** Interface.
**Physical source pages:** 47.
[PNG](<10_Experimental_Materials/P235-AppB-FigU06-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1022>) · Original paper and literature note are not hosted.
### P235 - App Fig. U07
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 7. **Visual forms:** Interface.
**Physical source pages:** 48.
[PNG](<10_Experimental_Materials/P235-AppB-FigU07-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1023>) · Original paper and literature note are not hosted.
### P235 - App Fig. U08
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 8. **Visual forms:** Interface.
**Physical source pages:** 49.
[PNG](<10_Experimental_Materials/P235-AppB-FigU08-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1024>) · Original paper and literature note are not hosted.
### P235 - App Fig. U09
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 9. **Visual forms:** Interface.
**Physical source pages:** 50.
[PNG](<10_Experimental_Materials/P235-AppB-FigU09-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1025>) · Original paper and literature note are not hosted.
### P235 - App Fig. U10
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 10. **Visual forms:** Interface.
**Physical source pages:** 51.
[PNG](<10_Experimental_Materials/P235-AppB-FigU10-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1026>) · Original paper and literature note are not hosted.
### P235 - App Fig. U11
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 11. **Visual forms:** Interface.
**Physical source pages:** 52.
[PNG](<10_Experimental_Materials/P235-AppB-FigU11-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1027>) · Original paper and literature note are not hosted.
### P235 - App Fig. U12
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 12. **Visual forms:** Interface.
**Physical source pages:** 53.
[PNG](<10_Experimental_Materials/P235-AppB-FigU12-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1028>) · Original paper and literature note are not hosted.
### P235 - App Fig. U13
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 13. **Visual forms:** Interface.
**Physical source pages:** 54.
[PNG](<10_Experimental_Materials/P235-AppB-FigU13-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1029>) · Original paper and literature note are not hosted.
### P235 - App Fig. U14
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 14. **Visual forms:** Interface.
**Physical source pages:** 55.
[PNG](<10_Experimental_Materials/P235-AppB-FigU14-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1030>) · Original paper and literature note are not hosted.
### P235 - App Fig. U15
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 15. **Visual forms:** Interface.
**Physical source pages:** 56.
[PNG](<10_Experimental_Materials/P235-AppB-FigU15-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1031>) · Original paper and literature note are not hosted.
### P235 - App Fig. U16
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 16. **Visual forms:** Interface.
**Physical source pages:** 57.
[PNG](<10_Experimental_Materials/P235-AppB-FigU16-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1032>) · Original paper and literature note are not hosted.
### P235 - App Fig. U17
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 17. **Visual forms:** Interface.
**Physical source pages:** 58.
[PNG](<10_Experimental_Materials/P235-AppB-FigU17-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1033>) · Original paper and literature note are not hosted.
### P235 - App Fig. U18
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 18. **Visual forms:** Interface.
**Physical source pages:** 58.
[PNG](<10_Experimental_Materials/P235-AppB-FigU18-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1034>) · Original paper and literature note are not hosted.
### P235 - App Fig. U19
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 19. **Visual forms:** Interface.
**Physical source pages:** 59.
[PNG](<10_Experimental_Materials/P235-AppB-FigU19-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1035>) · Original paper and literature note are not hosted.
### P235 - App Fig. U20
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 20. **Visual forms:** Interface.
**Physical source pages:** 59.
[PNG](<10_Experimental_Materials/P235-AppB-FigU20-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1036>) · Original paper and literature note are not hosted.
### P235 - App Fig. U21
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 21. **Visual forms:** Interface.
**Physical source pages:** 60.
[PNG](<10_Experimental_Materials/P235-AppB-FigU21-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1037>) · Original paper and literature note are not hosted.
### P235 - App Fig. U22
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 22. **Visual forms:** Interface.
**Physical source pages:** 61.
[PNG](<10_Experimental_Materials/P235-AppB-FigU22-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1038>) · Original paper and literature note are not hosted.
### P235 - App Fig. U23
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 23. **Visual forms:** Interface.
**Physical source pages:** 62.
[PNG](<10_Experimental_Materials/P235-AppB-FigU23-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1039>) · Original paper and literature note are not hosted.
### P235 - App Fig. U24
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 24. **Visual forms:** Interface.
**Physical source pages:** 63.
[PNG](<10_Experimental_Materials/P235-AppB-FigU24-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1040>) · Original paper and literature note are not hosted.
### P235 - App Fig. U25
**Paper:** Measuring Success in Education: The Role of Effort on the Test Itself
**Original caption:** No original caption supplied.
**Purpose:** Original computerized experimental test interface, Appendix B, Question 25. **Visual forms:** Interface.
**Physical source pages:** 64.
[PNG](<10_Experimental_Materials/P235-AppB-FigU25-Unnumbered%20research%20visual%20-%20no%20caption%20supplied%20in%20source.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1041>) · Original paper and literature note are not hosted.
### P242 - Fig. 2
**Paper:** Self-Regulated Learning: Beliefs, Techniques, and Illusions
**Original caption:** Figure 2. (Top panel) Exemplars from four bird families (materials from Wahlheim et al. 2011), with three exemplars (out of six studied) from the Jay family (from left to right: Green Jay, Blue Jay, Stellar Jay). (Bottom panel) Interface for choosing the next family. When a button for a given family is pressed, a bird (from the six presented in the familiarity phase) from the chosen family is presented for study along with the family name.
**Purpose:** Color-insert Figure 2 retains stimulus exemplars, study-choice interface and complete original caption. **Visual forms:** Interface, Composite.
**Physical source pages:** 29.
[PNG](<10_Experimental_Materials/P242-Fig02-(Top%20panel)%20Exemplars%20from%20four%20bird%20families%20(materials%20from%20Wahlheim%20et%20al.%202011),%20with%20three%20exemplars%20(out%20of%20six%20studied)%20from%20the%20Jay%20family%20(from%20left%20to%20right....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1042>) · Original paper and literature note are not hosted.
### P243 - Table 1
**Paper:** Socioeconomic gradients in children’s cognitive skills: are cross-country comparisons robust to who reports family background?
**Original caption:** Table 1: Questions on family background in PISA and PIRLS
**Purpose:** Recovered by direct visual comparison with source page. **Visual forms:** Complex table, Questionnaire comparison.
**Physical source pages:** 22.
[PNG](<10_Experimental_Materials/P243-Table01-Questions%20on%20family%20background%20in%20PISA%20and%20PIRLS.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1043>) · Original paper and literature note are not hosted.
### P252 - Fig. 1
**Paper:** The reliability paradox: Why robust cognitive tasks do not produce reliable individual differences
**Original caption:** Fig. 1 Schematic representation of tasks used and their conditions. Studies 1 and 2 featured the flanker, Stroop, go/no-go and stop-signal tasks. Study 3 featured the Posner cueing, SNARC and Navon tasks. Trials were presented intermixed in a randomized order. In the Go/no-go and Stop-signal tasks, visual stimuli were presented for a fixed duration of 1,250 ms (c.f. Verbruggen et al., 2008). In all other tasks, stimuli were presented until a response was given. An Inter-Stimulus Interval (ISI) of 750 ms was used in all tasks. Stimuli sizes are enlarged for illustration
**Purpose:** Shows stimuli and presentation rules for the seven tasks. **Visual forms:** Composite.
**Physical source pages:** 4.
[PNG](<10_Experimental_Materials/P252-Fig01-Schematic%20representation%20of%20tasks%20used%20and%20their%20conditions.%20Studies%201%20and%202%20featured%20the%20flanker,%20Stroop,%20go%20no-go%20and%20stop-signal%20tasks.%20Study%203%20featured%20the%20Posner....png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1044>) · Original paper and literature note are not hosted.
### P253 - App Table S1
**Paper:** Validity and fairness of the PISA 2018 Global Competence assessment: an argument-based evaluation via explanatory item response models
**Original caption:** Table S1. Item contents for measures
**Purpose:** Three-page framework mapping links merged/vertical global-competence dimensions to complete questionnaire item sets, index codes and scale reliabilities. **Visual forms:** Complex table, Questionnaire mapping.
**Physical source pages:** 2, 3, 4.
[PNG](<10_Experimental_Materials/P253-App-TableS1-Item%20contents%20for%20measures.png>) · [Source-quality PDF](<00_Source_Quality_Figures.pdf#page=1045>) · Original paper and literature note are not hosted.
## Exceptions and source limits
Missing original captions and original low-resolution images remain explicitly labelled in the manifest. Original raster information is preserved without generative enlargement. C085/C089 source papers were supplied as complete HTML; labelled local reading PDFs retain original source images and captions. DOCX appendices were exported through Microsoft Word on working copies. Older P161 visuals retain their existing embedded-image provenance. Long tables remain one logical table with ordered PNG parts. C082 contains only the identified dissertation-hosted article on physical pp. 23–34; other embedded studies are excluded.
## No-visual sources
- P063 / P063: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P075 / P075: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P087 / P087: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P098 / P098: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P106 / P106: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P112 / P112: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P125 / P125: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P138 / P138: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P148 / P148: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P165 / P165: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P168 / P168: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P184 / P184: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P186 / P186: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P187 / P187: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P194 / P194: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P161 / C021: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P232 / C030: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P209 / C042: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P208 / C156: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P225 / C160: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P246 / C264: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P220 / C266: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P229 / C272: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P077 / C016: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
- P246 / C263: no new qualifying visuals in this source, or duplicated visuals already retained in another indexed version.
## Unavailable or incomplete original captions
| Paper / visual | Physical pages | Source condition |
|---|---|---|
| [P124 - Fig. U09](#p124_p0023_v07) | 23 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P197 - Fig. U01](#p197_p0010_v05) | 10 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P124 - Fig. U01](#p124_p0015_v02) | 15 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P124 - Fig. U02](#p124_p0016_v02) | 16 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P124 - Fig. U03](#p124_p0017_v02) | 17 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P124 - Fig. U04](#p124_p0018_v02) | 18 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P124 - Fig. U05](#p124_p0019_v02) | 19 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P124 - Fig. U06](#p124_p0020_v03) | 20 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P124 - Fig. U07](#p124_p0021_v02) | 21 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P124 - Fig. U08](#p124_p0022_v02) | 22 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P195 - Fig. U01](#p195_p0017_v00) | 17 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P228 - Fig. 10](#p228_p0027_c15) | 27 | The supplied source truncates this caption; the available original wording is retained. |
| [P228 - Fig. 4](#p228_p0019_v01) | 19 | The supplied source truncates this caption; the available original wording is retained. |
| [P213 - App Fig. A1.2](#p213-app275_p0003_c00) | 3 | Original small raster notes beneath both panels retained in the crop; partly illegible in the supplied source even when rendered larger. No uncertain transcription invented. |
| [P253 - App Fig. S5](#p253-app278_p0020_c02) | 20 | The original caption and the plot internal title disagree; both are retained as printed. |
| [P253 - App Fig. S6](#p253-app278_p0020_c04) | 20 | The original caption and the plot internal title disagree; both are retained as printed. |
| [P235 - App Fig. U01](#p235_p0042_manual_material_q01) | 42 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U02](#p235_p0043_manual_material_q02) | 43 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U03](#p235_p0044_manual_material_q03) | 44 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U04](#p235_p0045_manual_material_q04) | 45 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U05](#p235_p0046_manual_material_q05) | 46 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U06](#p235_p0047_manual_material_q06) | 47 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U07](#p235_p0048_manual_material_q07) | 48 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U08](#p235_p0049_manual_material_q08) | 49 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U09](#p235_p0050_manual_material_q09) | 50 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U10](#p235_p0051_manual_material_q10) | 51 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U11](#p235_p0052_manual_material_q11) | 52 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U12](#p235_p0053_manual_material_q12) | 53 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U13](#p235_p0054_manual_material_q13) | 54 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U14](#p235_p0055_manual_material_q14) | 55 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U15](#p235_p0056_manual_material_q15) | 56 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U16](#p235_p0057_manual_material_q16) | 57 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U17](#p235_p0058_manual_material_q17) | 58 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U18](#p235_p0058_manual_material_q18) | 58 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U19](#p235_p0059_manual_material_q19) | 59 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U20](#p235_p0059_manual_material_q20) | 59 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U21](#p235_p0060_manual_material_q21) | 60 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U22](#p235_p0061_manual_material_q22) | 61 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U23](#p235_p0062_manual_material_q23) | 62 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U24](#p235_p0063_manual_material_q24) | 63 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
| [P235 - App Fig. U25](#p235_p0064_manual_material_q25) | 64 | Original caption unavailable or not printed; local extraction numbering does not invent an original figure caption. |
## Source-limited raster resolution
These originals contain limited-resolution raster content; rendering at a higher DPI does not create missing detail. Original pixels remain available in the source-quality album.
| Paper / visual | Physical pages |
|---|---|
| [P123 - Fig. 1](#p123_p0004_c05) | 4 |
| [P123 - Fig. 3](#p123_p0006_v02) | 6 |
| [P161 - Fig. 1](#p161_p0010_c14) | 11 |
| [P172 - Fig. 1a](#p172_p0019_c02) | 19 |
| [P161 - Fig. 3](#p161_p0019_c10) | 19 |
| [P163 - Fig. 1](#p163_p0034_c00) | 34 |
| [P068 - Fig. 1](#p068_p0009_c04) | 9 |
| [P161 - Fig. 2](#p161_p0017_c05) | 17 |
| [P068 - Fig. 2](#p068_p0014_c07) | 14 |
| [P123 - Fig. 6](#p123_p0021_v00) | 21 |
| [P143 - Fig. 2](#p143_p0009_c08) | 9 |
| [P172 - Fig. None](#p172_p0020_v01) | 20 |
| [P183 - Fig. 1](#p183_p0002_v02) | 2 |
| [P197 - Fig. 4](#p197_p0020_c03) | 20 |
| [P170 - Fig. 1](#p170_p0007_v04) | 7 |
| [P143 - Fig. 8](#p143_p0022_c03) | 22 |
| [P200 - Fig. 1](#p200_p0008_c03) | 8 |
| [P242 - Fig. 3](#p242_p0013_v05) | 13 |
| [P214 - Fig. 1](#p214_p0021_c09) | 21 |
| [P214 - Fig. 5](#p214_p0030_v00) | 30 |
| [P214 - Fig. 7](#p214_p0032_c08) | 32 |
| [P214 - Fig. 8](#p214_p0033_c01) | 33 |
| [P217 - Fig. 1](#p217_p0015_v01) | 15 |
| [P236 - Fig. 2](#p236_p0011_c05) | 11 |
| [P214 - Fig. 2](#p214_p0026_v01) | 26 |
| [P214 - Fig. 3](#p214_p0027_c07) | 27 |
| [P214 - Fig. 4](#p214_p0028_v01) | 28 |
| [P217 - Fig. 2](#p217_p0017_v02) | 16, 17 |
| [P236 - Fig. C1](#p236_p0025_v00) | 25 |
| [P236 - Fig. C2](#p236_p0026_v04) | 26 |
| [P243 - Fig. 2](#p243_p0028_c02) | 28 |
| [P243 - Fig. 3](#p243_p0029_c02) | 29 |
| [P243 - Fig. 4](#p243_p0030_c02) | 30 |
| [P253 - App Fig. S2](#p253-app278_p0007_c03) | 7 |
| [P235 - App Fig. U01](#p235_p0042_manual_material_q01) | 42 |
| [P235 - App Fig. U02](#p235_p0043_manual_material_q02) | 43 |
| [P235 - App Fig. U03](#p235_p0044_manual_material_q03) | 44 |
| [P235 - App Fig. U04](#p235_p0045_manual_material_q04) | 45 |
| [P235 - App Fig. U05](#p235_p0046_manual_material_q05) | 46 |
| [P235 - App Fig. U06](#p235_p0047_manual_material_q06) | 47 |
| [P235 - App Fig. U07](#p235_p0048_manual_material_q07) | 48 |
| [P235 - App Fig. U08](#p235_p0049_manual_material_q08) | 49 |
| [P235 - App Fig. U09](#p235_p0050_manual_material_q09) | 50 |
| [P235 - App Fig. U10](#p235_p0051_manual_material_q10) | 51 |
| [P235 - App Fig. U11](#p235_p0052_manual_material_q11) | 52 |
| [P235 - App Fig. U12](#p235_p0053_manual_material_q12) | 53 |
| [P235 - App Fig. U13](#p235_p0054_manual_material_q13) | 54 |
| [P235 - App Fig. U14](#p235_p0055_manual_material_q14) | 55 |
| [P235 - App Fig. U15](#p235_p0056_manual_material_q15) | 56 |
| [P235 - App Fig. U16](#p235_p0057_manual_material_q16) | 57 |
| [P235 - App Fig. U17](#p235_p0058_manual_material_q17) | 58 |
| [P235 - App Fig. U18](#p235_p0058_manual_material_q18) | 58 |
| [P235 - App Fig. U19](#p235_p0059_manual_material_q19) | 59 |
| [P235 - App Fig. U20](#p235_p0059_manual_material_q20) | 59 |
| [P235 - App Fig. U21](#p235_p0060_manual_material_q21) | 60 |
| [P235 - App Fig. U22](#p235_p0061_manual_material_q22) | 61 |
| [P235 - App Fig. U23](#p235_p0062_manual_material_q23) | 62 |
| [P235 - App Fig. U24](#p235_p0063_manual_material_q24) | 63 |
| [P235 - App Fig. U25](#p235_p0064_manual_material_q25) | 64 |
## Unnumbered original visuals
U identifiers are local extraction numbers. Any available original section heading or title is preserved separately.
| Paper / local visual | Physical pages |
|---|---|
| [P124 - Fig. U09](#p124_p0023_v07) | 23 |
| [P156 - Fig. U01](#p156_p0020_v03) | 20, 21 |
| [P197 - Fig. U01](#p197_p0010_v05) | 10 |
| [P124 - Fig. U01](#p124_p0015_v02) | 15 |
| [P124 - Fig. U02](#p124_p0016_v02) | 16 |
| [P124 - Fig. U03](#p124_p0017_v02) | 17 |
| [P124 - Fig. U04](#p124_p0018_v02) | 18 |
| [P124 - Fig. U05](#p124_p0019_v02) | 19 |
| [P124 - Fig. U06](#p124_p0020_v03) | 20 |
| [P124 - Fig. U07](#p124_p0021_v02) | 21 |
| [P124 - Fig. U08](#p124_p0022_v02) | 22 |
| [P195 - Fig. U01](#p195_p0017_v00) | 17 |
| [P205 - Table U01](#p205_p0029_v02) | 29, 30, 31 |
| [P205 - Table U02](#p205_p0031_v01) | 31, 32, 33, 34 |
| [P207 - Table U01](#p207_p0013_manual_rubric) | 13 |
| [P238 - App Table U01](#p238-app260_p0002_manualunnumbered1) | 2 |
| [P235 - App Fig. U01](#p235_p0042_manual_material_q01) | 42 |
| [P235 - App Fig. U02](#p235_p0043_manual_material_q02) | 43 |
| [P235 - App Fig. U03](#p235_p0044_manual_material_q03) | 44 |
| [P235 - App Fig. U04](#p235_p0045_manual_material_q04) | 45 |
| [P235 - App Fig. U05](#p235_p0046_manual_material_q05) | 46 |
| [P235 - App Fig. U06](#p235_p0047_manual_material_q06) | 47 |
| [P235 - App Fig. U07](#p235_p0048_manual_material_q07) | 48 |
| [P235 - App Fig. U08](#p235_p0049_manual_material_q08) | 49 |
| [P235 - App Fig. U09](#p235_p0050_manual_material_q09) | 50 |
| [P235 - App Fig. U10](#p235_p0051_manual_material_q10) | 51 |
| [P235 - App Fig. U11](#p235_p0052_manual_material_q11) | 52 |
| [P235 - App Fig. U12](#p235_p0053_manual_material_q12) | 53 |
| [P235 - App Fig. U13](#p235_p0054_manual_material_q13) | 54 |
| [P235 - App Fig. U14](#p235_p0055_manual_material_q14) | 55 |
| [P235 - App Fig. U15](#p235_p0056_manual_material_q15) | 56 |
| [P235 - App Fig. U16](#p235_p0057_manual_material_q16) | 57 |
| [P235 - App Fig. U17](#p235_p0058_manual_material_q17) | 58 |
| [P235 - App Fig. U18](#p235_p0058_manual_material_q18) | 58 |
| [P235 - App Fig. U19](#p235_p0059_manual_material_q19) | 59 |
| [P235 - App Fig. U20](#p235_p0059_manual_material_q20) | 59 |
| [P235 - App Fig. U21](#p235_p0060_manual_material_q21) | 60 |
| [P235 - App Fig. U22](#p235_p0061_manual_material_q22) | 61 |
| [P235 - App Fig. U23](#p235_p0062_manual_material_q23) | 62 |
| [P235 - App Fig. U24](#p235_p0063_manual_material_q24) | 63 |
| [P235 - App Fig. U25](#p235_p0064_manual_material_q25) | 64 |