Bias-corrected false discovery rates
Date: Wednesday, November 11, 2020
Time: 4:00 pm - 5:00 pm
Abstract: This talk explains why data analyses based on false discovery rates tend to be misleading due to a bias and shows how to correct that bias. Specifically, there is a bias estimating a false discovery rate with a significance level set to the value of the estimated false discovery rate. That bias leads to an excessive number of false positives, which are rejections of true null hypotheses such as hypotheses of no differential gene expression or hypotheses of no genetic association. Two methods designed to correct that bias are explained. Such bias correction can be interpreted as calibrating estimates of false discovery rates to become estimates of local false discovery rates. An exposition of this material appeared as chapter 6 of Genomics Data Analysis: False Discovery Rates and Empirical Bayes Methods (CRC, 2019).