Primary Submission Category: Heterogeneous Treatment Effects
Randomization-Based Inference for Average Treatment Effects in Inexactly Matched Observational Studies
Authors: Jianan Zhu, Jeffrey Zhang, Zijian Guo, Siyu Heng,
Presenting Author: Jianan Zhu*
Matching is a widely used causal inference study design in observational studies. Ideally, treated units are exactly matched with controls for the covariates, and randomization-based inference for the treatment effect can then be conducted as in a randomized experiment under the ignorability assumption. However, matching is typically inexact when continuous covariates or many covariates exist. Previous studies have routinely ignored inexact matching in the downstream randomization-based inference as long as some covariate balance criteria are satisfied. However, these inference methods focus on the constant treatment effect (i.e., Fisher’s sharp null) and are not directly applicable to the average treatment effect (i.e., Neyman’s weak null). To address this important gap, we propose a new framework — inverse post-matching probability weighting (IPPW) — for randomization-based inference for average treatment effects under inexact matching. Compared with the routinely used randomization-based inference framework based on the difference-in-means estimator for average treatment effects, our proposed IPPW framework can substantially reduce bias due to inexact matching and improve the coverage rate. Our framework can also be extended to the instrumental variable settings to simultaneously address the bias due to inexact matching and unmeasured confounding bias. We have also applied our framework to an observational study of kidney diseases among agricultural workers in Zimbabwe.