I have 2 groups of subject who underwent 4 conditions in an fMRI study. I
want to show where activation in these 4 conditions correlates with a
continuous variable. I have an average measure of the continuous variable
for each subject in each condition. I have been struggling to specify the
appropriate second-level design.
I chose 'full factorial" design and specified the scans representing
activity for each condition in each group. I specified group and condition
as factors (8 total factors). I also specified the 1 covariate.
The design does not produce the results I expected. I get a F tests for
1)avg effect of condition, 2)main effect of group,3) main effect for
condition, and 4) interaction between group and condition. However, the
main effect of the covariate is not shown in the results manager.
It seems that the authors of SPM have chosen not to report the F test of the
beta for the covariate in the results( although of course it has been
estimated). So if this is correct, what model should I be employing to
answer my question.
Is it multiple regression? If so, if so i need to clarify 2 things. 1)Do i
need 8 covariates with values for each relevant cell and zeros for the 7
other cells? Do i need to center the covariate across the 8 conditions ahead
of time?
Thank you for any help you can provide,
Jen
|