Fig. 2From: How to perform prespecified subgroup analyses when using propensity score methods in the case of imbalanced subgroupsBias according to the treatment effect in subset 1 (Scenario 2). Comparison of the “across subsets” or “within subsets” strategy in terms of the bias (A) variance (B) and the coverage of the 95% CI (C) for the estimation of the treatment effect in each subsetBack to article page