Debiased Bayesian Inference for High-dimensional Regression Models
Prof. Ruixuan LIU
Associate Professor, Department of Decisions, Operations and Technology
Associate Professor (by courtesy), Department of Economics
CUHK Business School
The Chinese University of Hong Kong
Date: 11 February 2026 (Wednesday)
Time: 10:30-12:00
Venue: E22-G015
Host: Prof. Degui LI, Distinguished Professor in Business Economics
Abstract
There has been significant development of sparsity-inducing priors, such as Spike-and-Slab or Horseshoe type priors, in the context of high-dimensional regression models. However, the resulting posterior does not generally possess desirable frequentist properties. The credible set does not form confidence set even asymptotically. In this paper, we introduce a novel approach that extends the concept of debiasing LASSO type estimators (Bayesian point estimators) to debias the entire Bayesian posterior distribution. We establish a new Bernstein-von Mises Theorem that guarantees the frequentist valid of the corrected posteriors. We demonstrate the practical performance of our proposal through Monte Carlo results and two real economics applications. This is based on joint with Qihui Chen (CUHK-SZ) and Zheng Fang (Emory).
Speaker
Prof. Ruixuan LIU is an Associate Professor at the Chinese University of Hong Kong (CUHK) Business School. His research is in the area of econometrics and data science. His recent works focus on nonparametric Bayesian inference and its applications to econometric models. His research work has been published on Econometrica, Journal of Econometrics, Econometric Theory and Quantitative Economics, etc. He won the 2018 Arnold Zellner Price (jointly with Yanqin Fan) for the best theoretical econometrics paper published by Journal of Econometrics between 2016 and 2017. He is currently an associate editor of Journal of Econometrics and Econometric Reviews.
All are welcome!
