A Comparison of the False Discovery Rate with Dunnett’s Test for a Large Number of Treatments

Abstract

Probabilities of two types of errors are estimated for the False Discovery Rate Procedure (FDR), Dunnett’s test, and conducting individual t-tests when treatments are being compared against a control. Probabilities for two types of powers are also estimated. The number of treatments considered is 50, 100, 150, and 300 with the first treatment being the control. Differing numbers of treatments equal to the control are considered with equal sample sizes of 3 and 5.