Vicky,
You are welcome. In your case, if the B condition is a real condition (as apposed
to "rest") I would model both conditions. You could enter the number of sessions
(# of subjects) and 2 conditions (A and B). If you have a true rest, enter 1
condition and allow SPM to use an implicit rest.
So, for 1 condition, your contrast would be [1 1 1 1 -1 -1 -1 -1] for a study with
4 subjects in each group. This would show the areas where activity is greater in
group 1 (the four subjects with a contrast of 1) than in group 2 (the four subjects
with a contrast of -1). Then, a contrast of [-1 -1 -1 -1 1 1 1 1] would show where
group 2 is greater than group 1.
If you have two true conditions and want to make the direct comparison A > B
(and/or B > A) the group contrast is more complicated.
To see where A>B you would need a contrast of [1 -1] for each subject. Then, for
the group contrast of 4 subjects in each group you would enter the following:
For group 1 > group 2 [1 -1 1 -1 1 -1 1 -1 -1 1 -1 1 -1 1 -1 1] or for group 2 >
group 1 [-1 1 -1 1 -1 1 -1 1 1 -1 1 -1 1 -1 1 -1].
However, I think that most SPM experts would recommend using random effects (two
sample t-test) to evaluate differences between groups.
I have also sent this to the list. So, if anyone else has any input I would be
interested in hearing what you think.
It can get confusing. I hope this helps,
Paul Laurienti
> Victoria Morgan wrote:
>
> > Paul,
> > Thanks for your help. I want to try the fixed effects comparison without
> > the second model. After I enter the number of sessions, it asks number of
> > conditions. If this is the A-B paradigm described, do I say one and then
> > proceed to model that? I thought then I would use the contrast [1,0] to see
> > the effect of A. Where does the session to session contrast [1 1 1 1 1 1
> > -1 -1 -1 -1 -1] come in?
> >
> > Thanks,
> > Vicky
> >
|