MerQuR+Statistics Glossary

Statistics Glossary → Levene's test

Levene's testconcept

A test of whether two or more groups have equal variances (spreads).

The absolute distance of each observation from its group mean is computed, and the mean distances are compared between groups. If the distances are clearly larger in one group, that group is more spread out. The null hypothesis is 'the variances are equal'; if p < α the variances are concluded to differ.

The version that uses the distance from the median instead of the mean is called the Brown-Forsythe test and is more robust for skewed data. Bartlett's test asks the same question but is very sensitive to departures from normality.

An everyday example

Two supermarket chains sell the same product. In one, the price barely varies from branch to branch; in the other, branches differ a lot. Levene's test looks not at the average price but at how large the variation between branches is.

In MerQur

In the assumption options of the t-test and ANOVA; in the t-test, the 'Automatic' variance choice uses Levene's test to decide between Student and Welch.

Analyses where this concept appears

Independent-Samples t-TestOne-Way ANOVA

Related concepts

Welch correction