【12.28腾讯会议】Robust High-Dimensional Regression with Coefficient Thresholding and its Application to Imaging Data Analysis
报告题目： Robust High-Dimensional Regression with Coefficient Thresholding and its Application to Imaging Data Analysis
报告学长: Lingzhou Xue（薛凌洲）
报告地点：腾讯会议 ID: 381 6014 5369
摘要：It is of importance to develop statistical techniques to analyze high-dimensional data in the presence of both complex dependence and possible outliers in real-world applications such as the imaging data analyses. We propose a new robust high-dimensional regression with coefficient thresholding, in which an efficient nonconvex estimation procedure is proposed through the thresholding function and robust Huber loss. The proposed regularization method accounts for complex dependence structures in predictors and is robust against outliers. Theoretically, we carefully analyze the landscape of the nonconvex loss function for the proposed method, which enables us to establish both statistical and computational consistency under the high-dimensional setting. Finite-sample properties of the proposed method are demonstrated by extensive simulation studies. An illustration of real-world application concerns a scalar-on-image regression analysis of the general factor of psychopathology based on task functional magnetic resonance imaging data from the Adolescent Brain Cognitive Development study.
报告人简介：Lingzhou Xue is an Associate Professor of Statistics at The Pennsylvania State University and Associate Director of the National Institute of Statistical Sciences. He received B.Sc. in Statistics from Peking University in 2008 and Ph.D. in Statistics from the University of Minnesota in 2012. He was a postdoctoral research associate at Princeton University from 2012-2013.
His research interests include high-dimensional statistics, statistical machine learning, optimization, econometrics, and statistical applications in biological science, business analytics, environmental science, and social science.