MAPSM: Stata module to perform multiple arms propensity score matching

The mapsm command; stands for “Multiple Arms Propensity Score Matching”, matches the propensity scores predicted from binary logistic regression (or multinomial logistic regression), which indicates the likeliness of choosing or being assigned to one among other treatment arms for each subject. In case of two-treatment arms study, the propensity score (probability of 0-1.0) derived from binary logistic regression will be divided into 10 strata, 0-0.1, 0.1-0.2 until 0.9-1.0. Study subjects from either of the two arms will be sampled to match with the opposite arm within the same propensity score strata, with 1:1 ratio. In case of three or more arms, the propensity scores will be derived from multinomial logistic regression, yielding sets of propensity scores equal to number of arms, indicating the likeliness of choosing or being assigned to one among other treatment arms for each subject. Each set of the propensity scores will be divided into 10 strata similar to the two-treatment arms. Subjects from each arm within the same strata of the propensity score will be sampled to obtain a matched set of 1:1:1 ratio in case of three arms, and 1:1:1:1 in case of four arms, and so on. In order to obtain the most similar post-matched contrast groups, the command also rerun the matching process and reported the best post-matched contrast groups which has the best balance diagnostic property, such as standardized mean difference. The best seed-setting number (among those selected) will be reported in order to re-obtain the best post-matched cohort.