American Review of Mathematics and Statistics

An Alternate Test for Variances
Gerry La Bute

Abstract
Current tests for variances have various limitations. Those such as the chi-square test for one variance or the F test for two variances require the assumption of normality. The Hartley test for multiple variances requires equal sample sizes as well as the assumption of normality. Levene’s test has the limitation that if the variance of one group increases relative to the variance of the other groups, the variation within groups can increase at a greater rate than the variation between groups resulting in the ironic situation of a smaller test statistic. The purpose of this paper is to present a nonparametric test for variances that can be used for any number of populations without the restriction of normality or equal sample sizes.

Full Text: PDF     DOI: 10.15640/arms.v4n1a2