@

초록

Average treatment effect on the treated (ATT) is a reasonable estimand in the causal inference problem with a rare-exposure regime. Although ATT estimation methodologies have been proposed and compared in previous studies, systematic experiment-comparison studies are lacking. In this work, we review and compare the performance of various ATT estimation methods, with focus on weighting-based methods and the rare exposure regime. Our numerical experiments demonstrate that a doubly robust-type estimator shows reasonable estimation performance in many cases. An use-case of the ATT estimators is illustrated with nationwide healthcare insurance claims data to estimate the effect of obesity on adverse events among the patients with Type-II diabetes.

키워드