Step-Down Multiple Comparison Procedures Using Medians and Permutation Tests

UNCG Author/Contributor (non-UNCG co-authors, if there are any, appear on document)
Scott J. Richter, Professor (Creator)
Institution
The University of North Carolina at Greensboro (UNCG )
Web Site: http://library.uncg.edu/

Abstract: Richter and McCann (2007) presented a median-based multiple comparison procedure for assessing evidence of group location differences. The sampling distribution was based on the permutation distribution of the maximum median difference among all pairs, and provides strong control of the FWE. This idea is extended to develop a step-down procedure for comparing group locations. The new step-down procedure exploits logical dependencies between pairwise hypotheses and provides greater power than the single-step procedure, while still maintaining strong FWE control. The new procedure can also be a more powerful alternative to existing methods based on means, especially for heavy-tailed distributions.

Additional Information

Publication
Communications in Statistics: Simulation and Computation, 38 (8), 1551-1561
Language: English
Date: 2009
Keywords
Median difference, Pairwise comparisons, Permutation test, Step-down procedure, 62G09

Email this document to