I have two groups with equal variances and unequal sample size (28 and 8) and I want to run a Student t-test. Can I use Mann-Whitney considering that the variances are equal? Do you think if there is a better test to compare the samples?
$\begingroup$
$\endgroup$
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
Theoretically, if the assumption of equal variances is satisfied and the dependent variable is normally distributed you could run a t-test despite the unequal sample sizes between the two comparison groups. I suggest you read these two posts (1 & 2) which deal with similar issues. If the assumption of normality is not satisfied then a Mann-Whitney test is appropriate as you have already suggested. You don't provide more information regarding your research questions and data availability to actually recommend a better statistical test.
$\begingroup$
$\endgroup$
Levene's test would allow you to compare the medians of the groups, the test does not require that the distribution is Gaussian, and it does not expect equal sample sizes (although it is preferred).
Instructions for calculating Levene's test in R can be found here and it can be calculated in a number of other programs. If the software you are using insists on equal sample size, go through the notes in the first link from NIST to perform the test manually.
