报告人简介

汪敏，美国德州大学圣安东尼奥分校 (University of Texas at San Antonio) 商学院管理科学与统计系副教授（获终身教职），博士生导师。2010年5月于美国克莱姆森大学（Clemson University）获得统计硕士学位；2013年5月于克莱姆森大学大学获得统计博士学位。2013年8月- 2017年12月在美国密歇根理工大学数学科学系工作和在2017年8月破格提前提升为副教授并获得终身任期教授资格；现在在德州大学圣安东尼奥分校从事教学科研工作。近年来，先后参与和主持了美国自然科学基金委（NSF），密歇根交通部，以及美国卫生院（NIH）的研究课题。其研究成果已发表在IISE Transactions, Naval Research Logistics, International Journal of Production Research, Bayesian Analysis, Computer & Industrial Engineering, The American Statistician, Computational Statistics & Data Analysis等国际权威期刊。主要研究方向包括贝叶斯统计；计算统计；统计推断；质量和可靠性工程研究；高维数据分析和统计应用。

Part 01

报告题目

Novel Bayesian Approaches for Variable Selection in Quantile Regression Models

Part 02

参会方式

腾讯会议 ID：284 629 215

2021年11月29日：9：00-10：30

Part 03

报告摘要

Asymmetric Laplace (AL) specification has become one of the ideal statistical models for Bayesian quantile regression, as it not only offers fast convergence of Markov Chain Monte Carlo (MCMC), but also guarantees posterior consistency under model misspecification. However, variable selection under such a specification is a daunting task because, realistically, prior specification of regression parameters should take the quantile levels into consideration. In this talk, we first develop a novel three-stage computational scheme for the recent proposed quantile-specific g-prior, which starts with an expectation-maximization algorithm, followed by Gibbs sampler and ends with an importance re-weighting step that improves the accuracy of approximation. We then consider the problems of parameter estimation and variable selection in Bayesian hierarchical quantile regression model in high-dimensional settings, in which the model dimension could greatly exceed the sample size. An efficient sampling algorithm based on Gibbs sampler and Metropolis-Hastings algorithm to draw samples from the full conditional posterior distributions to make posterior inference. Finally, the performance of the proposed methods is examined through simulation studies and real-data applications.

来源丨新媒体中心

编辑丨伍嘉怡