> I would very much appreciate your help on a question regarding the use of
> basic models on a VBM study: I am analyzing two sets of images aquired in
> distinct in two samples (longitudinal or within subjects design). I would
> like to use a covariate when comparing pre vs post gray matter volumes of
> these subjects. I see that in the t.paired test or the ANOVA (within
> subjects) desing there is not a possibility of including a covariate. I
> could then use the ANCOVA model but the problem is that the vector of my
> covariate has the same lenght as the number of images of the pre OR post
> conditions. This is to say, given that I have 8 subjects for the pre and
> the same 8 subjects for the post analyses, my covariate is also of 8
> values, one single value for each pair of images pre-post. Of course the
> ANCOVA model asks me for a 16-vector lenght of covariates (because I have
> included 8 pre and 8 post images) and I have had the same problem with
> other (fMRI and PET) models I have tried. I have also analyzed the data
> with an ANCOVA but repeating twice the vector of the covariate so that, for
> example the first and the ninth images corresponding to the first subject
> would have the same value in the covariate and so on. However I am not sure
> this approximation is appropiate.
>
> I would be grateful if someone could help resolving this issue.
This one may help:
http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0502&L=spm&P=R21538&I=-1
Your design matrix may look like the following, with the first two rows
corresponding to the pre and post scan of the first subject, the second pair
of rows being the pre and post scans of the second subject, etc. c1, c2, c3,
etc are the values of the covariate you are interested in.
0 1 0 0 ...
c1 1 0 0 ...
0 0 1 0 ...
c2 0 1 0 ...
0 0 0 1 ...
c3 0 0 1 ...
: : : :
A contrast of [-1 0 0 0 ...] or [1 0 0 0 ...] should tell you how the
differences covary with your covariate.
Best regards,
-John
|