1974
DOI: 10.1177/001316447403400406
|View full text |Cite
|
Sign up to set email alerts
|

An Empirical Comparison of the Anova F-Test, Normal Scores Test and Kruskal-Wallis Test Under Violation of Assumptions

Abstract: The present research compares the ANOVA F-test, the Kruskal-Wallis test, and the normal scores test in terms of empirical alpha and empirical power with samples from the normal distribution and two exponential distributions. Empirical evidence supports the use of the ANOVA F-test even under violation of assumptions when testing hypotheses about means. If the researcher is willing to test hypotheses about medians, the Kruskal-Wallis test was found to be competitive to the F-test. However, in the cases investiga… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

3
58
0
2

Year Published

1983
1983
2019
2019

Publication Types

Select...
8
1

Relationship

0
9

Authors

Journals

citations
Cited by 106 publications
(66 citation statements)
references
References 11 publications
(9 reference statements)
3
58
0
2
Order By: Relevance
“…If the data could not be transformed to normality, a GLMM was still used as recommended by Feir-Walsh and Toothaker, 1974 (Feir-Walsh and Toothaker, 1974). Post-hoc analyses utilized least squares means when the omnibus test was significant.…”
Section: Methodsmentioning
confidence: 99%
“…If the data could not be transformed to normality, a GLMM was still used as recommended by Feir-Walsh and Toothaker, 1974 (Feir-Walsh and Toothaker, 1974). Post-hoc analyses utilized least squares means when the omnibus test was significant.…”
Section: Methodsmentioning
confidence: 99%
“…There were some instances where transformations did not completely normalize the data; however, the GLMM was still performed. This was done because although the ANOVA F-test is strongest for normally distributed data, when comparing means of non-normally distributed data, the ANOVA F-test is still recommended [Feir-Walsh and Toothaker 1974]. Post hoc comparisons were done using t-tests.…”
Section: Methodsmentioning
confidence: 99%
“…Bradstreet (1997) found the rank transform test (Conover & Iman, 1982) to result in severely inflated Type I error rates. For the case of k > 2, Feir-Walsh and Toothaker (1974) and Keselman, Rogan, and FeirWalsh (1977) found the Kruskal-Wallis test (Kruskal & Wallis, 1952) and expected normal scores test (McSweeney & Penfield, 1969) to be "substantially affected by inhomogeneity of variance" (p. 220).…”
Section: Robustness With Respect To Unequal N's and Population Normalitymentioning
confidence: 99%