A Sequential Combined P-Value Test for Multiple Hypothesis Testing in Genomic Studies

• Activity Number: 253
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Shunpu Zhang*+ and Huann-Sheng Chen and Ruth Pfeiffer
• Companies: University of Nebraska and National Cancer Institute and National Cancer Institute
• Keywords: GWAS ; FWER ; type I error ; multiple testing

Abstract

Two combined p-value tests (the truncated product method (TPM) of Zaykin et al. (2002) and the rank truncated product (RTP) test of Dudbridge and Koeleman (2003) have been proposed, and are widely used in genome-wide association analysis. Instead of combining all the p-values, these tests combine only a subset of the p-values. It is claimed that these two tests are capable of providing a list of hypotheses in which there is at least one hypothesis that is truly false. An obvious drawback of the TPM and RTP tests is that the size of the subset of the p-values being included in the test statistics needs to be pre-selected subjectively. We propose a new step-up combined p-value test which does not require selecting the p-values for the test statistic, has weak control of the family-wise error rate (FWER), and can provide individual statements on hypotheses when the global null hypothesis is rejected. In simulations we demonstrate that the proposed test enjoys both the advantages of the combined p-value test and the classical multiple tests, and has the most robust performance to departures from the independent assumption among the existing combined p-value tests discussed in this pap