Simultaneous Multiple Comparisons with a Control Using Medians and Permutation Tests

• Activity Number: 424
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Scott Richter*+ and Melinda McCann
• Companies: The University of North Carolina at Greensboro and Oklahoma State University
• Keywords: Simultaneous inference ; multiple comparisons with a control ; median difference ; permutation test

Abstract

We investigate properties of simultaneous multiple comparisons with a control (MCC) procedures based on permutation tests and median statistics. The procedure for making all pairwise comparisons of Richter & McCann (2007) is modified for use in the case of all pairwise comparisons with a control. The procedure controls the familywise error rate by permuting only within pairs of treatments being compared, and retaining the maximum absolute difference among all pairs for the reference distribution. Simulation results suggest that the new procedure can have higher power than normal theory procedures when data come from heavy-tailed distributions.