Degtiar Rose 2023 - Generalizability and Transportability Review
Routing Summary
This folder fully captures Degtiar and Rose (2023), A Review of Generalizability and Transportability (Annual Review of Statistics and Its Application; arXiv 2102.11904): estimands (PATE, SATE), identifying assumptions, dissimilarity diagnostics, heterogeneity tests, weighting/matching, outcome-regression, doubly robust and combined-data estimators, recommendations and the summary-level appendix. 10 notes plus this index, with consolidated comparison tables and a reproducible marketing simulation.
- Need the paper on one page or the section-to-note map? → Degtiar Rose 2023 - Generalizability and Transportability Review - Overview
- Need PATE vs SATE, generalizability vs transportability, the sources of external validity bias? → Degtiar 2023 - Populations, Estimands and External Validity Bias
- Need the exact identifying assumptions (exchangeability, positivity, SUTVA for selection)? → Degtiar 2023 - Identifying Assumptions
- Need SMD, generalizability index, covariate shift, equivalence tests? → Degtiar 2023 - Assessing Dissimilarity Between Study and Target
- Need tests of treatment effect heterogeneity or what to do when generalizability fails? → Degtiar 2023 - Testing Treatment Effect Heterogeneity
- Need IPPW, odds weights, post-stratification, matching? → Degtiar 2023 - Weighting and Matching Methods
- Need outcome regression, BART, Bayesian evidence synthesis, bias-adjusted meta-analysis? → Degtiar 2023 - Outcome Regression Methods
- Need A-IPPW, TMLE, cross-design synthesis, combining RCT and observational data? → Degtiar 2023 - Doubly Robust and Combined Methods
- Need the recommendations, summary-level-data methods, or the method comparison tables? → Degtiar 2023 - Recommendations, Summary-Level Methods and Comparison Tables
- Need a runnable marketing simulation comparing the estimators? → Degtiar 2023 - Marketing Application Transporting Experiment Results
Concept Map
| Concept | Note | Type | Depends On | Key Result |
|---|---|---|---|---|
| PATE, SATE, PCATE, PATT | Degtiar 2023 - Populations, Estimands and External Validity Bias | definition | Causal Estimands | in target |
| External validity bias decomposition | Degtiar 2023 - Populations, Estimands and External Validity Bias | theorem | PATE, SATE | selection x heterogeneity x correlation |
| Internal validity assumptions A1-A3 | Degtiar 2023 - Identifying Assumptions | definition | Potential Outcomes Framework | exchangeability, positivity, SUTVA |
| External validity assumptions B1-B3 | Degtiar 2023 - Identifying Assumptions | definition | A1-A3 | , selection positivity, SUTVA |
| Generalizability index | Degtiar 2023 - Assessing Dissimilarity Between Study and Target | definition | selection propensity | ; <0.5 hard, >0.9 high |
| Tipton’s three paths | Degtiar 2023 - Testing Treatment Effect Heterogeneity | concept | dissimilarity | change target, state limits, new study |
| Post-stratification | Degtiar 2023 - Weighting and Matching Methods | definition | B1, B2 | |
| IPPW and odds weights | Degtiar 2023 - Weighting and Matching Methods | definition | selection propensity | for transport |
| Outcome regression standardization | Degtiar 2023 - Outcome Regression Methods | definition | B1 | marginalize fitted means over target |
| Hierarchical Bayesian evidence synthesis | Degtiar 2023 - Outcome Regression Methods | concept | bias term | adjusts unmeasured confounding in observational data |
| A-IPPW | Degtiar 2023 - Doubly Robust and Combined Methods | definition | IPPW, outcome regression | consistent if either model right |
| TMLE for encouragement designs | Degtiar 2023 - Doubly Robust and Combined Methods | definition | IV, EIC | clever covariate |
| Cross-design synthesis | Degtiar 2023 - Doubly Robust and Combined Methods | concept | RCT + observational | constant unmeasured confounding assumption |
| Seven recommendations | Degtiar 2023 - Recommendations, Summary-Level Methods and Comparison Tables | concept | all | define target, design, test, quantify, estimate, use all data, share code |
| Summary-level-data methods | Degtiar 2023 - Recommendations, Summary-Level Methods and Comparison Tables | concept | data availability | appendix table |
| Simulation: four estimators | Degtiar 2023 - Marketing Application Transporting Experiment Results | example | all estimators | PATE 0.6 vs naive 1.8 |
Notes
- Degtiar Rose 2023 - Generalizability and Transportability Review - Overview — CONTAINS: abstract; contributions; framework; Source Map; vocabulary table.
- Degtiar 2023 - Populations, Estimands and External Validity Bias — CONTAINS: Sec. 1 populations (Fig. 1), definitions, sources of bias, designs, Fig. 2 framework; Sec. 2 notation and estimands.
- Degtiar 2023 - Identifying Assumptions — CONTAINS: A1-A3, B1-B3 with weak forms; transportability modifications; correspondence with selection diagrams.
- Degtiar 2023 - Assessing Dissimilarity Between Study and Target — CONTAINS: Sec. 4.1 and 4.2 diagnostics table; generalizability index; Python.
- Degtiar 2023 - Testing Treatment Effect Heterogeneity — CONTAINS: Sec. 4.3 tests; Tipton’s three paths; comparison table; Python.
- Degtiar 2023 - Weighting and Matching Methods — CONTAINS: Sec. 5.1: matching, post-stratification, IPPW, odds weights, PS strata; summary table.
- Degtiar 2023 - Outcome Regression Methods — CONTAINS: Sec. 5.2: standardization, OLS, BART; multi-study Bayesian and meta-analytic methods.
- Degtiar 2023 - Doubly Robust and Combined Methods — CONTAINS: Sec. 5.3: TMLE, A-IPPW, calibration, domain adaptation, RCT+observational methods table.
- Degtiar 2023 - Recommendations, Summary-Level Methods and Comparison Tables — CONTAINS: Sec. 6 recommendations; Appendix; consolidated comparison tables; contrast with graphical approach.
- Degtiar 2023 - Marketing Application Transporting Experiment Results — CONTAINS: DGP, full code, single-run and 200-seed results, interpretation.
Sources
- Degtiar 2023 - A Review of Generalizability and Transportability — Degtiar I and Rose S (2023), Annual Review of Statistics and Its Application (arXiv 2102.11904v1), the sole source. The simulation and marketing mappings are interpretation.
See Also
- Pearl Bareinboim 2014 - External Validity and Transportability - Overview, Bareinboim Pearl 2016 - Causal Inference and Data Fusion - Overview — graphical identification counterparts.
- Potential Outcomes Framework, Causal Estimands, Bayesian Inverse Probability Weighting, Propensity Score Matching - Overview, Hierarchical Models, Geo-Experiment Methodology - Overview.