If you're still running the study, why don't you hold off on a group analysis? At only 3 and 7 now, the group results could end up changing dramatically. I think running a fixed-effects analysis takes up too much time to be worth it, at this stage.
________________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] On Behalf Of Abigail Livny-Ezer [[log in to unmask]]
Sent: Thursday, November 18, 2010 3:11 AM
To: [log in to unmask]
Subject: [SPM] Fixed effects analysis
Hi,
I have 2 groups of subjects I would like to compare between. In one group I have 3 subjects and in the other 7, (I am still running the study so eventually the groups will be more even and larger). I believe that in the meantime these groups are too small to run a random-effect model. So I am trying to apply a fixed-effect model, but didn't find in the manual (SPM8) any instructions or a data set examples.
I understand I am supposed to run the fixed-effects as if I am running again a 1st level analysis i.e. creating a new model specification in which I enter each subject as if he is a replication. Then, call for each subject's preprocessed images (swrs.img), enter the onsets and duration for each subject etc'. Then I estimate the model.
My design looks as if I have 7 replications one after the other in which there are 3 conditions.
1. I wonder if until now this was correct?
2. Also, I have a question regarding the contrast I should define. If I am interested in the 2nd condition> 1st condition, the contrast should be -1 1 0 or should I consider all 7 subjects so that the contrast will be -110 -110 -110 -110 -110 -110 -110 ?
3. Am I supposed to create 2 separate FFX group analyses- meaning one combining all 7 subjects and the other combining all 3 CTRL subjects?
4. How then do I compare between groups?
Thanks
Abigail
|