Covariate Balance as a Quality Indicator for Propensity Score Analysis

Propensity score analysis, such as propensity matching and weighting, is becoming increasingly popular in educational research. When a analysis conducted, examining the covariate balance considered to be crucial justify quality of results. However, it has been pointed out that solely considering how covariates after may not enough for justifying Suitable still yield biased estimates treatment effects. The current study aimed systematically demonstrate this problem by series simulation studies. As result, was revealed good on mean and/or variance does guarantee reduced bias an estimated effect. It also found estimation effect can unbiased some degree, even with lack under specific conditions.