Jerry,
I will take a stab at this.
I would recommend the random effects approach to test the interaction
between groups:
1. compute individual (first level) contrast images of A-B for each
subject:
2. Second level analysis: hit the basic models button.
select a two sample t-test
You dont need global normalization here since you did it (hopefully) at the
first level.
Enter the contrast images of A vs B in the young group for one sample and
then do same for older group.
3. The set up contrasts with Results button.
Your contrast should be:
1 -1 Young > Old
-1 1 Old > Young
Best,
Sterling C. Johnson, Ph.D.
Barrow Neurological Institute
222 W. Thomas Rd. Suite 315
Phoenix AZ 85013
-----Original Message-----
From: Jerry Allison [mailto:[log in to unmask]]
Sent: Friday, June 30, 2000 6:40 AM
To: [log in to unmask]
Subject: Group differences
I have two groups of fMRI subjects (10 young subjects and 10 old subjects).
Both groups performed an AB block design where B was a control task. All
data was spatially normalized. I have computed the effects of task A in each
individual subject, and for the youngsters as a group, and for the oldsters
as a group and for the total group.
How do I determine regions that were significantly more active in the young
group than in the old group (and vice versa).
Jerry Allison
Medical College of Georgia
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|