Fig. 2
From: Assessing racial disparities in healthcare expenditure using generalized propensity score weighting
Covariate balance under different ATT methods. Unadjusted: the simple difference in means of the two comparison groups without weighting; Conventional ATT: the original ATT estimation based on propensity score weighting but without handling positivity assumption violation; Trimming ATT: ATT estimation based on the subsample with propensity scores in [\(\alpha, 1\)], \(\alpha=0.10\); Truncation ATT: ATT estimation based on the entire sample but with propensity scores capped at \(\alpha=0.10\); Overlap Weighted ATT: ATT estimation based on the entire sample with individual-level overlap weights