Hello SPMers,
we are still struggling with a statistical/model problem: we'd like to test for differences
between 4 groups (sized 28, 24,24 and 20) at the '2nd level' (maps derived from
indendendent component analysis), however there are measurement repeats within groups
and between groups (e. g AABA ACDE BBCD CEDD...)
How could this be modelled in SPM5? Would a mixed model (fixed effect: for-level factor;
random effect: subject) be appropriate? How would it be set up & the subject ID's handled?
Or would the consideration of sphericity violations (unequal variance, non-independence) in
a normal ANOVA be sufficient?
As we are not stats experts we are grateful for any advice on how to handle this,
thanks very much in advance,
Philipp
|