A0303
Title: Inference of continuous treatment effects in large-scale observational data
Authors: Shujie Ma - University of California-Riverside (United States) [presenting]
Abstract: Recent advances in technology have created numerous large-scale datasets in observational studies, which provide unprecedented opportunities for evaluating the effectiveness of various treatments. Under the condition of unconfounded treatment assignment, most existing methods rely on a parametric or a nonparametric modeling method for estimating the propensity score or the outcome regression functions. The parametric approach lacks robustness as it suffers from the model misspecification problem. Conventional nonparametric estimation methods suffer from the curse of dimensionality when the dimension of confounders is large. A new method is introduced for estimating and inferring continuous treatment effects in large-scale observational data. The nuisance function is estimated by artificial neural networks. The proposed method takes full advantage of the large sample size of large-scale data and provides effective protection against misspecification bias. The theoretical properties established are presented for the proposed estimator, and the method is illustrated through simulation studies and real data applications.
