Performance of Tests for Homogeneity of Variances for More than Two Samples
To test for homogeneity of variances plays a critical role in not only conducting the analysis of variance (ANOVA)
but also monitoring the process in quality control. The performance of five tests including Cochran‟s test, Z-variance test,
O‟Brien test, Levene‟s test, and Modified Levene‟s Test was evaluated via a simulation study. Comparisons were made under
the null hypothesis and the alternative with various difference levels of variances. Overall, Cochran‟s test and Z-variance test
perform better than the other three tests and the two tests can be used in small sample sizes even five. Guidelines for selecting
tests for homogeneity of variances and alternative methods to compare treatment means in ANOVA when the homogeneity of
variances is violated are provided.
Keywords - Analysis Of Variance, Equality Of Variances, Heteroscedasticity, Test For Homogeneity Of Variances.