Dear FSLers,
I would like to compare an experimental group (n=19) to controls (n=23), using permutation methods.
I would typically use a simple two-sample t-test with randomise, but I would like to assume unequal variances (since groups are unbalanced, because experimental groups usually have a larger variance, and since I can see in my data that variances are quite different).
I thus think I should use PALM, but I cannot figure how to set up the exchangeability blocks and variance groups for this extremely basic design... (the whole point being to permute labels between groups, right?)
Could someone please help me? Is it that by default PALM assumes unequal variances while randomise assumes they are equal? Or do I need to specify these -eb and -vg options? What am I missing?
Thank you!
Best,
Florence
|