you should treat each run as a single session in the fixed effects
analysis - otherwise the intensity differences between runs will not be
properly modeled. I generally analyze each individual separately at the
fixed effects level because it's faster, but there are some people who
analyze them all at once. in either case, what you should do for random
effects is perform a contrast that averages across the three runs (e.g.,
if there is just one condition for three runs, then it would probably be
[1 1 1]). Then take the contrast image from that contrast for each
subject and put them into the random effects model. That will ensure
that you have the proper number of degrees of freedom for the random fx
model.
cheers
russ
wrote:
>An fMRI story---
>for each subject we performed three runs- each run was essentially
>the same except for the order of conditions (blocks) was
>counterbalanced. I used Roger Woods tools to do my data pre-
>processing and now have everything in 'analyze formate' -- I'm
>intending to eventually do a random effects analysis but understand I
>must first generate a fixed effects model --- QUESTION: Should I
>concatinate runs so that I have one long session for each subject or
>should I treat each run as a single session in my fixed effects model?
>
>
>(for random effects)
> If I treat each run as a single session ......
>
>I have 12 subjects - - I end up with 36 sessions - is there any way to
>create 'true' single subject contrast (incorporating the 3 runs) within
>the fixed effect model or do I creat a single contrast for each of the 36
>sessions that will then be used in the random effects analysis? If I
>end up with 36 single 'session' contrast' do I need to then specify in
>my random effects model that session 1,2,3 all correspond to subject
>one or do I just work with the 36 contrast as independent measures?
>
>Thank YOU - Thank YOU - THANK YOU - to whomever can help me
>resolve these issues----
>
>
|