Causal inference for quantile treatment effects

Analyses of environmental phenomena often are concerned with understanding unlikely events such as floods, heatwaves, droughts, or high concentrations pollutants. Yet the majority causal inference literature has focused on modeling means, rather than (possibly high) quantiles. We define a general estimator population quantile treatment (or exposure) effects (QTE)—the weighted QTE (WQTE)—of which is special case, along class balancing weights incorporating propensity score (PS). Asymptotic properties proposed WQTE estimators derived. further propose and compare PS regression two methods based these to understand effect an exposure quantiles, allowing for be binary, discrete, continuous. Finite sample behavior three studied in simulation. The applied data taken from Bavarian Danube catchment area estimate 95% phosphorus copper concentration river.