统计学系系列讲座之311期

时间：2018年6月21（周四）9:00-10:00

地点：史带楼204室

主持人：黄达 博士 复旦大学管理学院统计学系

主题：Pseudo estimation and variable selection in regression

主讲人：Professor Xiangrong Yin  University of Kentucky

简介：Xiangrong Yin is Professor of Statistics at the University of Kentucky since 2014. He obtained his PhD degree in 2000 at the University of Minnesota. He was assistant professor, associate professor and professor at the University of Georgia (2000-2014). His paper with his adviser R. D. Cook won the 2001 The Inaugural Editor's Award for the best article published in the Australian and New Zealand Journal of Statistics. His paper with his student Yuan Xue won The Journal of Nonparametric Statistics Best Student Paper Prize 2015. He was an associate editor for Statistica Sinica (2014-2017) and Statistics and Probability Letters (2010-2014. He has been an associate editor since 2010 for Journal of Nonparametric Statistics. He has guided twelve PhD students and his research interests are sufficient dimension reduction, multivariate analysis and big data analytics. He has published 59 papers, including JASA, JRSSB, Biometrika and AOS.

摘要：Often a problem in linear regression either for the correlated data or for the high-dimensional data is the singularity of the sample covariance matrix. While dealing with an ill-conditioned covariance matrix has been a longstanding challenge in the statistical literature, recent proposals relying on adding noises to the original data have found their successes in obtaining reliable estimates and making predictions. It has been shown that perturbing data with noises has a close relationship to many penalized estimators such as the well-known ridge estimator and lasso estimator. In this talk, we propose to add noises to the predictors with a known covariance structure, and call the estimator obtained in such a way a ``pseudo estimator''. A new variable selection procedure based on the concept of pseudo confidence interval which works for both the correlated predictors and the ``large p small n'' problem is proposed. We study the theoretical properties of the proposed pseudo estimator and variable selection procedure. Furthermore, we use an ensemble step to stabilize the pseudo estimation (variable selection) results. The accuracy and stability of the proposed method are demonstrated by both simulation studies and real data analyses.

                                                           统计学系 

2018-6-13