Hello,
I'm sure this is a foolish question, but I've been getting myself
confused. I've got data from an fMRI experiment with a standard 2x2
factorial event related design and I've modeled each subject's data
with an HRF and its derivative and dispersion in SPM2. So I get 12
beta images (2x2 factorial x 3) for each of my 20 subjects. Now I
want to go to the second level and find the main effects of my
factors. I don't have any specific hypotheses about temporal effects,
but I'm expecting to get a basic HRF-shaped response in all the
regions I'm interested in, and I need to know if this activation is
bigger in some regions than in others, using the most powerful
analysis I possible.
So should I ...
a) Calculate contrasts for each subject for each main effect with
each basis function and then do something with them ... ??
b) Ignore my deriv and dispersion columns and just do the main
effects using t-tests on the HRFs?
c) Take all 12 betas for each subject into a big ANOVA and do some
kind of F test? I tried getting a main effect of A with:
hAB hAb haB hab dAB dAb daB dab sAB sAb saB sab
1 1 -1 -1 0 0 0 0 0 0 0 0
0 0 0 0 1 1 -1 -1 0 0 0 0
0 0 0 0 0 0 0 0 1 1 -1 -1
where h = hrf, d = deriv, s = dispersion, and A and B are my factors
But here I'm not sure if I need to this as a repeated measures ANOVA
or just a straight forward one.
Methods b and c give fairly similar results, but I don't know which is
right or why. I guess the real question is are my Deriv and Disp
columns a useful signal in themselves, or should I consider them as
regressors of non-interest and focus just on my HRF column.
Can anyone comment?
Antonia.
|