PSTRATA: Stata module for optimal propensity score stratification

pstrata stratifies the propensity score into strata (quantile-based bins) chosen so that, within each stratum, the treatment groups are balanced on the propensity score. In general, stratification (also referred to as subclassification) allows the investigator to analyze outcomes between treatment groups within each stratum as well as to observe overall differences between groups across all strata (Linden & Adams 2008). It has been shown that stratification of the propensity score into five quantiles can remove over 90% of the initial bias due to the covariates used to create the propensity score (Cochran 1968; Rosenbaum & Rubin 1984). While the strata generated by pstrata can be used in any subsequent analysis, it was specifically intended for use in conjunction with mmws; which can be downloaded from SSC. Before using pstrata, the propensity score for a given treatment must be estimated. A logit or probit regression model can be used for estimating the propensity score for a binary treatment and a multinomial logistic or probit regression can be used for estimating propensity scores for multiple treatment levels. pstrata then generates one or more strata variables (named strata1, strata2, etc.), depending on the number of treatment groups in the study.