It is remarkable that the assumptions of the Kruskal-Wallis are milder than those of the F
test, but it is important to note that the less extensive are the assumptions for performing a
test, the more general will be its conclusions and the lower its efficiency. Thus, if the underlying
assumptions of the analysis of variance are met, the F test will present better performance than
the nonparametric Kruskal-Wallis (CAMPOS, 1983).