Estimating heterogenous treatment effect with big data

What is it about?

In the big data era, it is important to get robust estimation for potentially heterogenous treatment effect with many variables available to the researcher. In this paper we propose new nonparametric estimators for the reduced dimensional conditional average treatment effect (CATE) function. In the first stage, we use machine learning methods to select relevant variables, allowing the number of covariates to be comparable to or larger than the sample size. The second stage consists of a low-dimensional local linear regression, reducing CATE to a function of the covariate(s) of interest. We derive functional limit theory for the estimators and provide an easy-to-implement procedure for uniform inference based on the multiplier bootstrap.

Featured Image

Why is it important?

The method is robust to potential model misspecifications and could handle high-dimensional data. We also propose inference procedures for the unknown functions of the treatment effect contingent on the characteristics of the individuals subject to treatment/intervention.