A Two-stage Method for Multiple-hypothesis-testing

• Activity Number: 423
• Type: Topic Contributed
• Date/Time: Thursday, August 12, 2004 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Shuhui Wen*+ and Chu-Hsing Hsiao
• Companies: National Taiwan University and Yuanpei University of Science and Technology and National Taiwan University
• Keywords: association test ; Bonferroni ; false positive rate ; SNP marker ; multiple-hypothesis-testing ; two-stage

Abstract

Recent association studies often involve a large number of markers under investigation and thus lead to the problem of multiple hypothesis testing. Traditional Bonferroni's procedure is too conservative in the sense that much attention has been focused on reducing the Type I error and a strict threshold is employed due to a large number of tests considered simultaneously. The current interest, however, lies in the identification of markers mostly of mild effect. In other words, the goal may be better achieved via reducing the probability of false negatives among those with association. The two types of errors cannot be minimized at the same time. We propose a two-stage method for multiple testing procedure. The main principle is to remove those obviously unassociated markers and retain those of mild effect in the first stage, followed by a test that adjusts for the possibly incurred error. We provide the implementation of the proposed procedure and derive its theoretical success and failure rates. We also illustrate the derivation of the sample size under this design and recommend the choice between two stages. The performance of the two-stage method is evaluated.