It is shown that the undesirable properties of the rank transform in two-way analysis of variance with an interaction are due to the test construction and not to the sampling distribution used. It is furthermore confirmed that a safe procedure is to test for main effects only if the interaction is shown not to be significant.
Get full access to this article
View all access options for this article.
References
1.
LemmerH. H. (1980) Some empirical results on the two-way analysis of variance by ranks. Communications in Statistics—Simulation and Computation, 14, 1427–1438.
2.
SawilowskyS. S. (2000) Review of the rank transform in designed experiments. Perceptual and Motor Skills, 90, 489–497.
