I have been attempting to determine the most appropriate design to use for a
second-level analysis for determining group differences in the correlation
between a behavioral score and the brain.
From thoroughly reading the archives as well as available information on
SPM5 designs, it seems that choosing design=multiple regression with 3
covariates 1) behavioral score, 2) indicator variable for group 1, and 3)
indicator variable for group 2 is recommended followed by t-contrasts for
group, e.g., 0 1 -1 0 and 0 -1 1 0.
The problem with the above model seems to be the inability to look at the
"simple effects", i.e., the estimated regions with "significant" positive
betas for group 1, estimated areas with "significant" negative betas for
group1....etc.
I would have thought I could specify 0 1 0 0, 0 0 1 0, 0 -1 0 0, 0 0 -1 0
but this simply is not allowed by SPM.
At any rate, this seems to be a fairly straightforward analysis question:
Are the group differences in the correlation. If there are group
differences, what are the within group betas? These with group betas are the
only way to describe the reason for the differences.
Any help, input, would be greatly appreciated. i realized i may have phrased
this question in many ways recently but this one last attempt to get an
answer. If i am missing something obvious that everyone should know please
point me in the correct direction. I greatly need a push in the correct
direction.
|