Implications of Extreme P-Values for Power and Replication in Large-Scale Genomic Studies

• Activity Number: 23
• Type: Topic Contributed
• Date/Time: Sunday, July 29, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Government Statistics
• Author(s): Laura Lazzeroni*+ and Ilana Belitskaya-Levy and Ying Lu and Amrita Ray
• Companies: Stanford University and VA Palo Alto Cooperative Studies Program Coordinating Center and Stanford University and Stanford University
• Keywords: multiple testing ; p-values ; GWAS ; simultaneous inference ; power ; replication

Abstract

Large-scale genome-wide association studies rely on extremely small p-values on the order of 10^-8 to achieve positive findings. Little attention has been given to the implications of this shift from the traditional 0.05 critical value on the likely success of such studies. Increasing array sizes and the adoption of genome sequencing and other next generation technologies will only exacerbate this problem. In this talk, we address the following questions. 1. What are the practical limits for the number of hypothesis tests in a genomic study, given the constraints of feasible sample sizes and plausible effect sizes? 2. What does an extreme p-value really say about the relative strength of a "genome-wide significant" finding relative to a more modest one? How likely is such a finding to be replicated? We obtain general results applicable to most statistical analyses that provide explicit and somewhat unexpected answers to these questions.