Pairwise testing of group mean vectors in MANOVA with small samples |
| |
| Authors: | Philip H Ramsey Patricia P Ramsey |
| |
| Institution: | 1. Department of Psychology , Queens College of CUNY , 65-30 Kissena Blvd., Flushing , NY , 11367 , USA phramsey@aol.com;3. Management Systems , Fordham University Graduate School of Business Administration , 113 West 60th Street, New York , NY , 10023 , USA |
| |
| Abstract: | Several, multivariate, pairwise, multiple comparison procedures are proposed as follow-ups for a significant multivariate analysis of variance. The Peritz procedure is generalized from univariate to several multivariate applications. Procedures are evaluated using overall power, any-pair power and all-pairs power applied to mean vectors with common sample sizes of 4, 5, and 9. Monte Carlo simulation demonstrated greater power than previously proposed univariate procedures in many conditions especially for all-pairs power. The multivariate Peritz procedure based on the Lawley–Hotelling trace was found to be most powerful in many conditions. |
| |
| Keywords: | Pillai–Bartlett trace greatest characteristic root Wilks’ likelihood ratio Hotelling–Lawley trace Peritz procedure |
|
|
