I I want to perform a paired t-test to check if there's some effect, I have the distribution of "before" and the distribution of "after" the manipulation. Do I need to assume the population variances of the two distribution are equal ?

  • 4
    $\begingroup$ The paired t-test is a t-test applied to the differences, so the distribution of the differences is what matters. $\endgroup$
    – whuber
    Commented May 31, 2023 at 16:31

1 Answer 1


Any assumptions you make in a paired test would have to do with the paired differences. After all, a paired test is a one-sample test in disguise.

Therefore, NO, you do not need to assume equal variances of the two groups for a paired t-test.

  • $\begingroup$ The paired $t$-test considers whether the mean has changed: it cannot spot whether or not the variance has changed. $\endgroup$
    – Henry
    Commented Jun 1, 2023 at 9:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.