PRACTICAL CONSIDERATIONS IN CHOOSING A MANOVA TEST STATISTIC - A REJOINDER TO STEVENS

The recommendation of the Pillai-Bartlett trace V for general use as the test statistic in multivariate ANOVA is reexamined and reaffirmed in light of criticisms by J. P. Stevens (see record 1979-25171-001). Empirical data are presented show that the V test's substantial robustness advantage does not require extreme violations of assumptions, and examples are cited to show the occurrence in practice of the type of heteroscedasticity in which the V test particularly excels and the occurrence of the pattern of population mean differences for which the V test is more powerful than its rivals. (10 ref) (PsycINFO Database Record (c) 2006 APA, all rights reserved).