Speaker: Xintao Xia, Graduate Assistant

Title: Multiply Robust Inference of Treatment Effects under High-dimensional Clustered Data

Abstract: In this paper, we develop a multiply robust inference procedure for the average treatment effect (ATE) under high-dimensional clustered data and the ignorable assumption. We consider the case where the target population is formed from heterogeneous sources with different treatment selection mechanisms, and it is difficult to correctly specify a single parametric model for such propensity scores (PS). We propose to apply multiple clustering algorithms and fit working PS models on each estimated cluster. This provides a set of estimated PS models. We develop a novel high-dimensional empirical likelihood weighting method under soft covariate balancing constraints to combine multiple working PS models and a regularized augmented outcome regression to correct the bias due to non-exact covariate balancing. The proposed confidence interval for the ATE has valid asymptotically nominal coverage under high-dimensional covariates if any of the PS models, their linear combination, or the outcome regression (OR) model is correctly specified. We demonstrate the advantages of the proposed approach over the existing doubly robust inference methods under high-dimensional covariates via both theoretical and numerical results. We analyzed the right heart catheterization (RHC) dataset, initially collected from five medical centers and two different phases of studies, to demonstrate the utility of the proposed method in practice.