A New Method for Variable Selection in Global Fréchet Regression
Holder： Changcheng Li（Dalian University of Technology）
Time：2025-03-27 14:00-15:00
Location：Siyuan Lecture Hall，Zhi Hua Building-225
Abstract:
Studying the regression relationship between response variables and predictor variables is an important topic in statistics. To address the scenario where the response variable lies in a general metric space, global Fréchet regression is developed, and Fréchet ridge selection operator(FRiSO) is introduced to achieve variable selection for global Fréchet regression. However, in cases where predictor variables are collinear, FRiSO no longer accurately accomplishes variable selection, and the computational efficiency of FRiSO is not satisfactory. To deal with this issue, we propose a new method that not only accomplishes variable selection accurately but also boosts computational efficiency for collinear predictor variables. We establish theorems and conduct numerical studies to demonstrate the variable selection capability of our method.
About the Speaker:
李长城，大连理工大学数学科学学院教授、博士生导师，研究兴趣主要为高维统计、高维因果推断及因果图学习。高维统计及因果推断是目前统计领域的热点，在智能决策、生物医学等诸多领域里都有非常重要的应用。在高维统计的理论、应用以及计算方面进行了一系列创新性的研究，在国际顶尖学术期刊Journal of American Statistical Association、Annals of Statistics、Journal of Econometrics、Annals of Applied Statistics等发表文章多篇。入选小米青年学者、辽宁省及国家级青年人才项目。
Your participation is warmly welcomed!
欢迎扫码关注北大统计科学中心公众号，了解更多讲座信息!
