I there,
ok you were not far .. ;-)
you need the multiple regression model witout cste (prob access via
spm_spm_ui rather than the GUI since I never managed to have no cste
otherwise .. )
1st reg. the group belonging say for 2*3 subjects (hope you have more
though .. ): 111-1-1-1
2nd regressor: behavioural scores group one: X X X 0 0 0
3rd regressor: behavioural scores group two: 0 0 0 X X X
other regressors = confound, e.g. age, and you have to input the total
grey matter
contrasts:
1 0 0 0 0 = group 1 > group 2
-1 0 0 0 0 = group 1 < group 2
0 1 0 0 0 = + corr in group 1
0 0 1 0 0 = + corr in group 2
0 1 -1 0 0 = corr in group 1 > corr in group 2
---> it is important to 1) see what;s happening in each group and 2)
plot the data --> you may for instance have corr gp1 > corr gp2 because
gp2 have no corr vs. a smaller corr ..
gd luck
cyril
> 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.
>
>
>
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
|