Online Program Home
My Program

Abstract Details

Activity Number: 358 - Contributed Poster Presentations: Section on Statistics in Epidemiology
Type: Contributed
Date/Time: Tuesday, July 30, 2019 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistics in Epidemiology
Abstract #307352
Title: Controlling the False Discovery Proportion: a Simulation Study
Author(s): HARLAN MCCAFFERY* and Chi Chang
Companies: University of Michigan and Michigan State University
Keywords: False Discovery Rate; multiple testing; simulation; R; hypothesis testing

In multiple testing situations, procedures to control the False Discovery Rate (FDR) often used as a less stringent, more powerful alternative to procedures that control the Familywise Error rate (FWER). We use simulations to show how the FDR-controlling procedures control the proportion of false discoveries, Q, and its expected value, the FDR, particularly well when the proportion of hypotheses tested in which the null is true is moderate to large. Going beyond the expected value of Q (the FDR), we investigate the distribution of Q under different experimental parameters after using FDR-controlling procedures. The results show that the variance of Q is positively correlated with the proportion of true null hypotheses and with the control level. This suggests that FDR control procedures are best suited to moderate proportions of null hypotheses and lower control levels.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program