Hi,
I'm interested in comparing connectivity in a number of different networks before and
after a drug challenge. I've done a group ICA (temporal concatenation) in the pre and
post conditions across all subjects, and used template matching to pick the equivalent
networks between conditions. What I was thinking of doing next was directly comparing
the pre and post betas (and spatial patterns) for each network in each subject; i.e.
extract the timecourses for each spatial pattern in each subject in each condition, sort
appropriately, construct a design matrix, and run a GLM, comparing pre and post within
subjects, and then across the group.
1) How are the timecourses laid out in the tXX.txt files? There are N subjects, and N+1
timecourses, so I assume either the first or last timecourse is the average (the one that
gets graphed). I think it is the first one, followed by the individual subject timecourses in
the order of their appearance in the fsf file, but I'd like to confirm that. My individual
subject GLM's look peculiar (as if I were using the timecourse from the wrong subject)
and I'd like to know if I'm extracting timecourses incorrectly, or if there is some other
issue with my procedure.
2) Is there some conceptual problem with doing this? Since the pre and post spatial
patterns are derived separately, and then paired based on spatial similarity (and only
pairs that pass a similarity threshold are included), it should be valid to do the
comparison in the GLM as described, right?
Thanks,
Blaise
|