统计与数据科学学院
-------------------------------------------------
学术讲座
讲座题目：Local False Discovery Rate Based Approach to Grouped Hypotheses Testing
讲座人：Sanat K. Sarkar
Chair and Cyrus H. K. Curtis Professor
Department of Statistical Science
Fox School of Business
Temple University
时间： 7月6日上午10:30-11:30
地点：统计与数据科学学院432教室
Abstract: For simultaneous testing of multiple hypotheses grouped into a one-classified form, a novel framework involving a simple extension of the standard two-class mixture model from single to multiple groups is proposed. This newer framework provides a decomposition of each hypothesis-specific local false discovery rate (Lfdr) into conditional Lfdr for the hypothesis given that it is within a significant group and Lfdr for the group itself. An Lfdr based approach to controlling false discoveries across the entire collection of hypotheses and possessing an optimal property is produced in its oracle form under this framework. It is a powerful alternative to ignoring the group structure by simply pooling all the hypotheses into a single group and using commonly used Lfdr based method under the standard single-group two-class mixture model.Numerical studies show that our proposed method is indeed more powerful than its relevant competitors, at least in their oracle forms, in commonly occurring practical scenarios.A possible extension of this method from one- to two-way classified hypotheses using a framework with the hidden states of the hypotheses being modeled in terms of a Bernoulli matrix is also presented.
邀请人： 邹长亮教授
欢迎广大师生参加！
统计与数据科学学院
2018年6月26日