The propensity score allows nonrandomized data to mimic features of data from randomized controlled trials. Based on the propensity score, the distribution of baseline covariates are similar between treated and untreated subsamples. That is the method of propensity-score matching forms sets of treated and untreated subsamples that share similar propensity scores so that these data are like randomized data. The most prevalent propensity score matching is pair matching, wherein pairs (i.e., one-to-one) of treated and untreated subsamples are formed to have similar propensity scores.