When Sample-Splitting Choices Matter in TMLE and AIPW: Cross-Fitting, Targeting, and Repetition
Variants of targeted maximum likelihood estimation (TMLE) and augmented inverse-probability weighting (AIPW) with machine-learning nuisance models differ in which observations train them, how bias is corrected, and whether splits are repeated.
ProofPaper ↗
Key points
- In known-truth simulations of a binary-outcome average treatment effect, eight headline estimators shared each setting's library, inner folds and truncation bound; coverage differences were judged against a prespecified two-percentage-point margin.
- With one random forest on the primary data-generating mechanism, estimators fitting the propensity in-sample undercovered severely (0.12-0.48), whereas cross-fitting both nuisances raised coverage to 0.91-1.00, two overcovering.
- Repetition reduced partition dependence but made several intervals more conservative.
- Results support cross-fitting both nuisances when empirical-process conditions cannot be justified, and assessing truncation and AIPW-versus-TMLE sensitivity under poor overlap.
Sources (1)
- [1]When Sample-Splitting Choices Matter in TMLE and AIPW: Cross-Fitting, Targeting, and RepetitionarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 12, 05:17 AM
Variants of targeted maximum likelihood estimation (TMLE) and augmented inverse-probability weighting (AIPW) with machine-learning nuisance models differ in which observations train them, how bias is corrected, and whether splits are repeated.
In known-truth simulations of a binary-outcome average treatment effect, eight headline estimators shared each setting's library, inner folds and truncation bound; coverage differences were judged against a prespecified two-percentage-point margin.
Extractive summary: sentences quoted from the sources.