Hi, all. I can't quite figure out how to do this, but it seems like it
should be something FSL can do. Perhaps someone can give me a pointer?
I have 18 subjects, each of whom has 3 runs. I used an FIR model at
the first levels (5 basis functions per condition), then collapsed
across the three runs at the second level (fixed effects), so now I
have one gfeat per subject, each of which has a COPE for each basis
function in each condition.
I'd like to analyze the condition f-tests at the group level, now, but
I don't know how. My naive assumption was that I'd set up the
contrasts like this example with 2 subjects, and 5 basis functions:
1 0 1 0 0 0 0
1 0 0 1 0 0 0
1 0 0 0 1 0 0
1 0 0 0 0 1 0
1 0 0 0 0 0 1
0 1 1 0 0 0 0
0 1 0 1 0 0 0
0 1 0 0 1 0 0
0 1 0 0 0 1 0
0 1 0 0 0 0 1
The first two columns are the subject EVs, and the last five columns
show the basis functions.
I'd hoped to set up the contrasts like this:
0 0 1 0 0 0 0
0 0 0 1 0 0 0
0 0 0 0 1 0 0
0 0 0 0 0 1 0
0 0 0 0 0 0 1
So that I'd get one COPE for each basis function, and then I could
just run an f-test across them all to see where my group activations
for that condition were. From this, I could get estimated HRFs by
merging the first 5 COPEs, and so on.
Of course, the problem here is that in my EV matrix, the sum of the
first 2 columns is the sum of the last 5, which makes the stats
grouchy.
How do I do this?
Thanks!
Todd
p.s. Oddly enough, SPM doesn't seem to mind this setup at the higher
levels, which I thought was a basic problem with linear algebra. Any
idea how it avoids this problem?
|