Dear Michael,
I think this can be done with a full factorial design, but not in way you
described below.
This is what you want to:
1. Select full factorial and define a factor "group" with 3 levels
2. Select the GM images in the respective cell in each factor level. Don't
forget to assign the correct factor level each cell.
3. Define one covariate "score"
3.a. enter ALL (!) scores in a single vector under "values"
3.b. Under interactions, choose "with Factor 1"
3.c. optional: Choose one of the centering options
The latter option will split the the covariate according to the groups and
you can test for the effects that you describe. Please verify that the
scores are actually assigned to the correct GM image. You can do this, if
you press "Review", select the SPM.mat that you just created and choose
"Explore/Files and Factors" from the Design menu in the lower left window.
I recommend doing this because I actually obtained different mappings
whether I entered the covariate values as a row or column vector.
Good Luck,
Jan
Glabus, Michael wrote:
> This one's for the stats experts.
>
> I'm doing a 3 group VBM analysis and want to explore the correlations
> between GM x "score", and their interactions.
>
> Could this be addressed in spm5 with a full factorial model?
>
> I imagine entering GM regressors and 3 "score" regressors thus:
> [GM1 GM2 GM3 ScoreG1 ScoreG2 ScoreG3]
>
> and exploring main effects with:
>
> [0 0 0 1 1 1] positive and [0 0 0 -1 -1 -1] negative correlation between
> "score" and GM in all three groups.
>
> and interactions (looking for different regression slopes, Beta) with
>
> [0 0 0 1 -0.5 -0.5] Beta 1 > Beta2 , Beta3, etc.
>
> Does this make sense?
>
> Happy Holidays!
>
> Regards - Mike
>
--
Jan Gläscher, Ph.D. Caltech Brain Imaging Center
+1 (626) 395-4976 (office) Caltech, Broad Center, M/C 114-96
+1 (626) 395-2000 (fax) 1200 California Blvd
[log in to unmask] Pasadena, CA 91125
|