Alexander Giessing, Princeton University

Inference on Heterogeneous Quantile Treatment Effects via Rank-Score Balancing
Date
Nov 3, 2020, 11:00 am12:30 pm
Event Description

Understanding treatment effect heterogeneity in observational studies is of great importance for many scientific fields simply because the same treatment may affect different individuals differently. Quantile regression provides a natural framework for modeling such heterogeneity. In this talk, we propose a new method for inference on heterogeneous quantile treatment effects in the presence of high-dimensional covariates. Our method is informed by considerations in causal inference as well as high-dimensional quantile regression. Specifically, our estimator combines a debiased L1-penalized regression adjustment with a quantile-specific covariate balancing scheme. We present a comprehensive study of the theoretical properties of this estimator, including weak convergence of the heterogeneous quantile treatment effect process to a mixture of independent, centered Gaussian processes. Notably, the proposed estimator is semi-parametric efficient when the dimension is fixed. These results are based on new maximal and deviation inequalities for suprema of unbounded empirical processes, which are of independent interest. We illustrate the finite-sample performance of our approach through Monte Carlo experiments and an empirical example, dealing with the differential effect of mothers’ education on infant birth weight.

(Joint work with Jingshen Wang, Division of Biostatistics, UC Berkeley)

Biography: Alexander Giessing is a Postdoctoral Research Associate in the Department of Operations Research and Financial Engineering at Princeton University. He holds a PhD in Statistics from the University of Michigan, Ann Arbor, a MSc in Econometrics and Mathematical Economics from the London School of Economics and Political Science, and a bachelor’s degree from the University of Bonn. Alexander’s research interests include bootstrap and resampling techniques, causal inference, and quantile regression. The primary goal of his research is to develop statistically efficient and computationally robust methods for inference on high-dimensional data.

Recording