Combining Inverse Propensity Score Weighting and Boosting Algorithms to Estimate Causal Moderation Effects (Poster 4)

• In Event: Innovations in Statistical Power, Intervention Planning and Design, and Causal Effects

Abstract

This study introduces the Inverse Propensity Score Weighting Boosted Moderation Estimator (IBME), for estimating causal moderation effects with binary treatments and continuous moderators in observational studies. IBME addresses a critical methodological gap in educational research by combining inverse propensity score weighting with boosting algorithms. This approach enables researchers to model complex, non-linear relationships between treatments, moderators, and outcomes more accurately than traditional methods. Monte Carlo simulations demonstrate IBME's superior performance in reducing bias and improving efficiency compared to conventional approaches. The method significantly enhances covariate balance and shows robustness to complex relationships between variables. A case study applying IBME to prekindergarten effects on math achievement across socioeconomic status reveals non-linear relationships that were not apparent using traditional regression methods.

Authors

• Sirui Ren, University of North Carolina - Chapel Hill

• Puyao Ge, Rice University

• Nianbo Dong, University of North Carolina - Chapel Hill