RAND Statistics Seminar Series

Large-scale Simultaneous Hypothesis Testing

Presented by Professor Brad Efron
Department of Statistics
Stanford University
Thursday, September 18, 2003 4:00 pm
Main Conference Room

Abstract

"Multiple comparisons" used to mean doing 2 or 3 or 10 hypothesis tests at the same time. Now statisticians need to consider 500 or 5000 testing situations simultaneously, in microarray or spectroscopy analyses for example. New problems and new opportunities arise in these contexts. I will discuss an empirical Bayes approach, related to false discovery rates, focusing on the choice of an appropriate null hypothesis for large-scale hypothesis testing. Two genomics examples will be used to motivate the methodology.