opposite group. The KBM takes the weighted average of all the nonparticipants
and then matches, while RM is a variant of the NNM.
Hence the results for the NNM are presented first. Normally all PSM estimators
should yield the same results because with growing sample
size all become closer to comparing only exact matches.While nearest
neighbor matching, kernel based matching and radius matching are
only based on the propensity score, mahalanobis metric matching is a
multivariate covariate matching method that allows for inclusion of
the propensity score as well as other covariates. The choice of the
matching method usually involves a trade-off between average
matching quality (bias) and variance. In the current analysis the results
for the nearest neighbormatching are presented first as itmatcheswith
the nearest neighbor in the opposite group.