Treatment effect and double robust estimator at the quantiles

Abstract: The analysis of treatment effect at various quantiles for two or more treatment conditions is discussed. Treatment effects are estimated using (i) inverse propensity score weights; (ii) unconditional outcome distribution within each group. Through (i) and (ii) the standard double robust estimator is extended to evaluate the treatment effects of binary/multiple treatment options not only on average but also in the tails (quantiles), to assess whether the treatment effect is constant or varies across the quantiles.