MerQuR+Statistics Glossary

Statistics Glossary → Folded F test (variance ratio)

Folded F test (variance ratio)concept

A test of whether two groups have equal variances, based on the ratio of the larger variance to the smaller one.

The statistic is F' = larger variance / smaller variance, so the ratio is always 1 or more (the distribution is 'folded'). The two-sided p value is 2 · P(F > F') from the F distribution whose degrees of freedom are the n - 1 values of the two groups. How far the ratio is from 1 also shows directly how large the difference in spread is.

The test is very sensitive to departures from normality: with heavy-tailed distributions it often signals a difference even when the variances are equal. As a precondition for the t-test, Levene's or the Brown-Forsythe test is therefore more robust, and it is often best simply to use the Welch t-test.

An everyday example

Temperatures measured 20 times each by two thermometers in the same room are compared. The variance of one thermometer's readings is 0.09 and the other's is 0.36, giving F' = 4. A ratio this large suggests the second thermometer is clearly less consistent.

In MerQur

The 'Variance ratio, folded F' checkbox of the Independent t-Test; the variance ratio between groups also features in Propensity Score Analysis.

Analyses where this concept appears

Independent-Samples t-TestPropensity Score Analysis

Related concepts

Levene's testBartlett's test of equal variances