I have a general neuroimaging stats question that has been on my mind.
I’m using fsl’s randomise to compare 3 groups (this is a vbm analysis but I suppose the question is relevant to regressions in general), and although 2 of my groups are equal (n=28 each), my third group’s n is 97. I’m reading that most neuroimaging packages use type III ss estimation, which makes me think unbalanced groups do not need to be further addressed via weighting.
Am I overlooking something here? Do people typically use covariates to account for major differences in sample sizes across groups?