On Wednesday 19 October Egil Ferkingstad, PhD student at CIGENE and the Dept. of biostatistics, University of Oslo, will give a talk.
 Egil Ferkingstad |
Place: Meeting room lower level, Animal Science Building, UMB
False discovery rate (fdr) methodology plays an important role in situations where a large number of related hypotheses are tested simultaneously. In this work-in-progress talk, I show how Efron's empirical Bayes local fdr methods can be extended to the case where a covariate may influence the prior probabilities of null hypotheses. This can be done using what we have called covariate-modulated fdr. The new method is applied to analysis of expression quantitative trait loci (eQTL) data, by viewing heritability as the covariate influencing the prior probability of linkage. The method is applied to real data from an eQTL study, where it is illustrated that the use of the covariate-modulated fdr yields a substantial increase in power compared to using Efron's local fdr method.
