I have two groups, each with 6 observations. One group was generated by an old test method. The other group was generated by a new test method. I would like to test if the new method produces more consistent results.

My first instinct is a test for equal variances, with an alternative hypothesis that the new test has a lower variance. I'd like to do this one sided test at the .05 level.

Because my data doesn't look normal (looked at Q-Q plots for each group, for what that is worth with 6 points), I don't want to use the F-Test for unequal variances.

I've read that the Levene and Brown-Forsythe tests are more robust. But the test statistics are derived from absolute values of differences with the mean (Levene) or median (Brown-Forsythe). With an F test the test statistic will be greater (than 1) when the variance in the numerator is higher and between 1 and 0 when the variance in the numerator is lower. So a one-sided interpretation seems intuitive here. The Levene and BF tests will just give a higher test statistic when the variances are further apart and a lower test statistic when they are similar. So I don't understand how to test my one-sided hypothesis with these tests.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.