Dear SPM experts,
I read through some older threads about variance settings in Flexible Factorial ANOVAs, and noticed that some set the variance to "equal" for within-subject effects and others leave it at "unequal". This might be attributed to different assumptions maybe (?). Imagine one factor with two levels, an easy one and a difficult one (e.g. some working memory task). I would naively expect larger variability and thus larger variance for the difficult condition.
Now I wonder, would it be ok to set the variance to "unequal" per default? As far as I can see, this seems to be the standard assumption in GLM Flex (IN.EqualVar = [0 0] for both 2x2 Within and 2x2 Between).
Would it be ok to set the variance for "subjects" to "unequal" as well or should it be left at "equal"?
Thanks,
Gabor
|