Regression-adjusted estimation of quantile treatment effects under covariate-adaptive randomizations

Datasets from field experiments with covariate-adaptive randomizations (CARs) usually contain extra covariates in addition to the strata indicators. We propose incorporate these additional via auxiliary regressions estimation and inference of unconditional quantile treatment effects (QTEs) under CARs. establish consistency limit distribution regression-adjusted QTE estimator prove that use multiplier bootstrap is non-conservative The regression may be estimated parametrically, nonparametrically, or regularization when data are high-dimensional. Even misspecified, proposed inferential procedure still achieves nominal rejection probability null. When correctly specified, minimum asymptotic variance. also discuss forms adjustments can improve efficiency estimators. finite sample performance new methods studied simulations, an empirical application a well-known dataset concerned expanding access basic bank accounts on savings reported.