Best Practices of Double Score Matching for Estimating Causal Effects

• Activity Number: 166 - Non-Clinical Statistics, Personalized Medicine, and Other Topics
• Type: Contributed
• Date/Time: Tuesday, August 10, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Yunshu Zhang* and Shu Yang and Wendy Ye and Douglas Faries and Ilya Lipkovich and Zbigniew Kadziola
• Companies: NC State University and North Carolina State University and Eli Lilly and Company and Eli Lilly and Company and Eli Lilly and Eli Lilly and Company
• Keywords: Average treatment effect on the treated; Causal inference; Double robustness; Propensity score; Prognostic score

Abstract

Unlike in randomized clinical trials (RCTs), confounding control is critical for estimating the causal effects from observational studies. Matching methods are popular because they can be used to emulate an RCT hidden in the observational study. To ensure the key assumption hold, it is common to collect ample possible confounders, rendering dimension reduction imperative in matching. Three matching schemes based on the propensity score (PSM), prognostic score (PGM), and double score (DSM, the collection of the first two scores) have been proposed in the literature. However, it lacks a comprehensive comparison among the matching schemes and has not made inroads into the best practices including variable selection, choice of caliper and replacement. In this article, we characterize the statistical and numerical properties of PSM, PGM, and DSM. We show that DSM performs favorably with, if not better than, PSM and PGM. In particular, DSM is doubly robust in the sense that the matching estimator is consistent requiring either the PS or PG model is correctly specified. We also provide detailed instructions for DSM and illustrate the recommendations with comprehensive simulation studies.