Estimating Generalized Propensity Scores with Survey and Attrition Weighted Data
We extend prior work and show with analytic results that using survey sampling or attrition weights in the GPS estimation stage and the outcome model stage for continuous treatments can result in a more robust estimator than one that does not.