Answers are grounded in this paper and cite their sources.
In completely randomized experimental designs where population variances are equal under the null hypothesis, it is not uncommon to have multiplicative treatment effects that produce unequal variances under the alternative hypothesis. Permutation procedures are presented to test for (a) median location and scale shifts, (b) scale shifts only, and (c) mean location shifts only. Corresponding multivariate extensions are provided. Location-shift power comparisons between the parametric Bartlett-Nanda-Pillai trace test and three alternative multivariate permutation tests for five bivariate distributions are included.
No takes yet. Share an insight, caveat, or question.
Mielke et al. (1994) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: