报告时间：2018年10月25日上午10点
报告地点：博识楼434室
学术报告1
Title：Response Best Subset Selection Model and Estimation in Multivariate Linear Regression
Author: Jianhua Hu
Abstract: Numerous studies have been conducted on situations where response variables are given and only predictors are subject to variable selection. In practice, however, the number of responses that are truly depend on the predictors is not known prior to data analysis, and thus there is also a need for response variable selection, a topic on which limited research has been done. In this paper, we consider the problem of response variable selection, propose a novel response best subset selection model, and provide an efficient estimation procedure for performing response selection and regression coefficient estimation by introducing a penalty function to responses. Our estimation has the oracle properties: model consistency and asymptotic normality of the regression coefficient estimators. Our finite sample simulation experiments demonstrate that the developed model and procedure are efficient and promising. We thus apply this approach to study a real dataset.
报告人简介：胡建华，2000年获中南大学理学博士学位，2007年获加拿大温莎大学统计学博士学位。现为上海财经大学经济创新平台海外引进教师，数据科学与统计研究院特任研究员，博士生导师。对高维数据变量选择、空间统计模型、局部平稳时间序列以及多元分析等有理论和应用意义的科学问题，开展了若干极有价值的研究工作。主持包括国家自然科学基金面上项目在内的科研项目十多项。在包括《Statistica Sinica》、《Bernoulli》和《Science China Mathematics》等在内的国际国内权威期刊发表论文近四十多篇。
报告地点：博识楼434室
学术报告1
Title：Response Best Subset Selection Model and Estimation in Multivariate Linear Regression
Author: Jianhua Hu
Abstract: Numerous studies have been conducted on situations where response variables are given and only predictors are subject to variable selection. In practice, however, the number of responses that are truly depend on the predictors is not known prior to data analysis, and thus there is also a need for response variable selection, a topic on which limited research has been done. In this paper, we consider the problem of response variable selection, propose a novel response best subset selection model, and provide an efficient estimation procedure for performing response selection and regression coefficient estimation by introducing a penalty function to responses. Our estimation has the oracle properties: model consistency and asymptotic normality of the regression coefficient estimators. Our finite sample simulation experiments demonstrate that the developed model and procedure are efficient and promising. We thus apply this approach to study a real dataset.
报告人简介：胡建华，2000年获中南大学理学博士学位，2007年获加拿大温莎大学统计学博士学位。现为上海财经大学经济创新平台海外引进教师，数据科学与统计研究院特任研究员，博士生导师。对高维数据变量选择、空间统计模型、局部平稳时间序列以及多元分析等有理论和应用意义的科学问题，开展了若干极有价值的研究工作。主持包括国家自然科学基金面上项目在内的科研项目十多项。在包括《Statistica Sinica》、《Bernoulli》和《Science China Mathematics》等在内的国际国内权威期刊发表论文近四十多篇。