Dimension Reduction for Conditional Density Estimation with Applications to High-Dimensional Causal Inference

Speaker: Dr Fu Ouyang

Affiliation: The University of Queensland

Location: Level 6 Boardroom (629), Colin Clark Building (#39), St Lucia Campus

Zoom: https://uqz.zoom.us/j/82603079317

Abstract: We propose a novel and computationally efficient approach for nonparametric conditional density estimation in high-dimensional settings that achieves dimension reduction without imposing restrictive distributional or functional form assumptions. Our method introduces a sparsity condition, assuming only a small subset of covariates significantly influences the outcome. We develop an innovative conditional dependence measure and a modified cross-validation procedure to efficiently identify relevant covariates in a data-driven manner, thereby circumventing the need for subjective threshold selection. We demonstrate the practical utility of our dimension-reduced conditional density estimation by applying it to doubly robust estimators for average treatment effects. Notably, our proposed procedure can not only select relevant variables for nonparametric propensity score estimation but also inherently reduce the dimensionality of outcome regression variables through a refined ignorability condition. We evaluate the finite-sample properties of our approach through comprehensive simulation studies and an empirical study on the effects of 401(k) eligibility on saving using SIPP data.