False Discovery Control in Gaussian Models with Misspecified Covariance

• Activity Number: 685
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Ye Liang* and Joshua Habiger and Xiaoyi Min
• Companies: Oklahoma State University and Oklahoma State University and Yale School of Public Health
• Keywords: False discovery rate ; Bayesian multiple testing ; Zellner's g prior ; Spatial dependence

Abstract

False discovery control in multiple testing problems is challenging when data are dependent. We focus on the performance of the local false discovery rate (LFDR), also known as a Bayesian posterior probability, under the situation that the covariance structure is misspecified for Gaussian models. The paper reveals how the mission of FDR control is affected by the misspecification and includes various examples for demonstration.