CMStatistics 2015: Start Registration
View Submission - CMStatistics
B1289
Topic: Contributed on Multiple tests I Title: A discrete modification of the Benjamini-Hochberg procedure using a weighting approach Authors:  Sebastian Doehler - Darmstadt University of Applied Science (Germany) [presenting]
Abstract: In many multiple testing problems control of the false discovery rate is desired. This can be achieved by the Benjamini-Hochberg (BH) procedure when the $p-$values satisfy certain dependence properties. It is also known that the BH procedure controls the FDR if the $p-$values are discrete and e.g. independent. We take discreteness into account by using a weighted version of the BH procedure. We investigate the power of this method for empirical and simulated data.