Vol: 43 Page: 73

Authors:
Magdalena Ćwiklińska
Zofia Hanusz

Title:
 Power of parametric and nonparametric tests of several mean populations

Language: English

Keywords:
normality
Kruskal-Wallis test
the F test
power

Summary:
One of the main statistical problem is testing the hypothesis about equality of means of several populations. In practice, to test this hypothesis the F test is used. However, the F test needs normality of populations and homogeneity of variances. In practice, generally these assumptions are not fulfilled. In such cases nonparametric tests should be used. Practitioners apply the F-test without verifying assumptions, with confidence that this test is the best one. In the paper we show in simulation study that nonparametric Kruskal-Wallis test is not worse than the F test by mean of their sample significance levels, and power under several alternative conditions.

Display