-----------------------------------------------------------

Statistics Seminar Series

Thursday, Apr. 11, 2013, 4:00 PM
Cullimore Lecture Hall 2
New Jersey Institute of Technology

-----------------------------------------------------------

 

On Procedures Controlling the FDR for Testing Hierarchically

Ordered Hypotheses

 

Gavin Lynch

 

Department of Mathematical Sciences

New Jersey Institute of Technology

 

Abstract

 

Complex large-scale studies, such as those related to microarray and quantitative trait loci, often involve testing multiple hierarchically ordered hypotheses. However, most existing false discovery rate (FDR) controlling procedures do not exploit inherent hierarchical structure among the tested hypotheses. In this talk, I present theoretical developments toward controlling the FDR when testing the hierarchically ordered hypotheses. First, I offer a general framework under which hierarchical testing procedures can be developed. Then, I present hierarchical testing procedures which control the FDR under various forms of dependence. Simulation studies show that these proposed methods can be more powerful than alternative methods.