@article{zhao_zhang_yang_2022, title={Double score matching in observational studies with multi-level treatments}, volume={8}, ISSN={["1532-4141"]}, DOI={10.1080/03610918.2022.2118778}, abstractNote={While weighting methods are popular for comparing the effects of multi-level treatment in observational studies, their performance can be unstable in the presence of extreme values of the generalized propensity score (GPS). Matching methods are more resistant to GPS outliers but bear the risk of GPS model misspecification. In this article, we propose a double score matching (DSM) estimator of the pairwise average treatment effect (ATE) based on the GPS and the generalized prognostic score (GPGS) evaluated at one treatment level at a time. The de-biased DSM estimator not only maintains the advantage of matching methods but also alleviates the model dependence problem due to its double robustness: it consistently estimates the true pairwise ATE if either the GPS or the GPGS is correctly specified.}, journal={COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION}, author={Zhao, Honghe and Zhang, Xiaofei and Yang, Shu}, year={2022}, month={Aug} }
 @article{zhao_yang_2022, title={Outcome-adjusted balance measure for generalized propensity score model selection}, volume={221}, ISSN={["1873-1171"]}, DOI={10.1016/j.jspi.2022.04.004}, abstractNote={In this article, we propose the outcome-adjusted balance measure to perform model selection for the generalized propensity score (GPS), which serves as an essential component in estimation of the pairwise average treatment effects (ATEs) in observational studies with more than two treatment levels. The primary goal of the balance measure is to identify the GPS model specification such that the resulting ATE estimator is consistent and efficient. Following recent empirical and theoretical evidence, we establish that the optimal GPS model should only include covariates related to the outcomes. Given a collection of candidate GPS models, the outcome-adjusted balance measure imputes all baseline covariates by matching on each candidate model, and selects the model that minimizes a weighted sum of absolute mean differences between the imputed and original values of the covariates. The weights are defined to leverage the covariate–outcome relationship, so that GPS models without optimal variable selection are penalized. Under appropriate assumptions, we show that the outcome-adjusted balance measure consistently selects the optimal GPS model, so that the resulting GPS matching estimator is asymptotically normal and efficient. We compare its finite sample performance with existing measures in a simulation study. We apply the proposed method to two real data applications.}, journal={JOURNAL OF STATISTICAL PLANNING AND INFERENCE}, author={Zhao, Honghe and Yang, Shu}, year={2022}, month={Dec}, pages={188–200} }