学术报告第201402期

报告题目：The Control of the False Discovery Rate under Structured  Hypotheses

报 告 人： 郭文革 教授（美国新泽西工学院）

报告时间： 2014年1月9日15：30

报告地点： 南一楼中311会议室

邀 请 方：“多谱信息处理技术”国家级重点实验室

报告摘要：

The hypotheses in many multiple testing problems have  some natural structure based on prior knowledge such as Gene Ontology in gene  expression data. However, few false discovery rate (FDR) controlling procedures  take advantage of this natural structure. In this talk, we introduce new FDR  controlling procedures which account for the structural information of the  tested hypotheses. The first structure we examine is when all hypotheses have  been ordered beforehand. We firstly develop conventional fixed sequence FDR  controlling procedures which stop on the first acceptance. Then, we extend the  method and develop procedures which stop on the kth acceptance. Simulation  studies and real data analysis show that the newly developed procedures can be a  powerful alternative to the existing Benjamini-Hochberg and Benjamini-Yekutieli  procedures. If time is allowed, we discuss the testing of hierarchically ordered  hypotheses where hypotheses are arranged in a tree-like structure and introduce  new hierarchical FDR controlling procedures under different dependence  configurations.

报告人简介：

Wenge Guo is currently an Assistant Professor of  Statistics in the Department of Mathematical Sciences at the New Jersey  Institute of Technology. Before coming to New Jersey, he worked as a Research  Fellow at the National Institute of Environmental Health Sciences for two years.  he received his Ph.D. in System Engineering and Biostatistics from Huazhong  University of Science and Technology and the University of Cincinnati,  respectively. His research interests include: Large-scale multiple testing,  High-dimensional data analysis, Bioinformatics, Machine learning, Statistical  methods for clinical trials.