GPMatch: A Bayesian Doubly Robust Approach to Causal Inference with Gaussian Process Covariance Function As a Matching Tool

Abstract: Gaussian process (GP) covariance function is proposed as a matching tool in GPMatch within a full Bayesian framework under relatively weaker causal assumptions. The matching is accomplished by utilizing GP prior covariance function to define matching distance. We show that GPMatch provides a doubly robust estimate of the averaged treatment effect (ATE) much like the G-estimation, the ATE is correctly estimated when either conditions are satisfied: 1) the GP mean function correctly specifies potential outcome (Y^{(0)}); or 2) the GP covariance function correctly specifies matching structure. Simulation studies were carried out without assuming any known matching structure nor functional form of the outcomes. The results demonstrate that GPMatch enjoys well calibrated frequentist properties, and outperforms many widely used methods including Bayesian Additive Regression Trees. The case study compares effectiveness of early aggressive use of biological medication in treating children with newly diagnosed Juvenile Idiopathic Arthritis, using data extracted from electronic medical records.