Characterizing the effect of matching using linear propensity score methods with normal distributions

摘要: SUMMARY Matched sampling is a standard technique for controlling bias in observational studies due to specific covariates. Since Rosenbaum & Rubin (1983), multivariate matching methods based on estimated propensity scores have been used with increasing frequency medical, educational, and sociological applications. We obtain analytic expressions the effect of using linear score normal distributions. These cover cases where either known, or discriminant analysis logistic regression, as typically done current practice. The results show that not only reduces along population score, but also controls variation components orthogonal it. Matching rather than can therefore lead relatively large variance reduction, much factor two common settings close matches are possible. Approximations given magnitude this which be computed estimates obtained from pools. Related reduction presented suggest that, difficult situations, use leads greater scores.