Dear Esa,
> If the selected contrasts for a conjunction analysis are not orthogonal,
> SPM asks the orthogonalization order of contrasts.
> The results seems to depend on which contrast is first in
> orthogonalization order. The first contrast remains the same, but those
> who are non-orthogonal regarding the first contrast change.
One assumption made for a conjunction analysis in SPM99 is that the used
contrasts are orthogonal in such a way that the spaces spanned by the
contrasts are orthogonal to each other, i.e. if X is your design matrix
and c1 and c2 two contrasts, then orthogonalizing c2 with respect to c1
means that c2 is changed such that
c1'*pinv(X) is orthogonal to c2'*pinv(X)
where pinv(X) = (X'*X)^(-1)*X'
>
> How to interpret the results of an orthogonalized conjunction analysis?
I don't know of an generally applicable interpretation about
orthogonalized conjunction analysis, but why not just say that you
modified the contrasts such that the spaces spanned by the contrasts are
orthogonal to each other. This removes common subspaces spanned by more
than one contrast from all but one contrast, which would otherwise make
your conjunction analysis invalid. Maybe others might want to comment
here.
>
> By the way, I already have articles "A test for a conjunction" by
> Worsley and Friston and "Multisubject fMRI studies and Conjunction
> analyses" by Friston et al. Are there other references concerning SPM99
> style conjunctions?
Not that I'm aware of...
Stefan
--
Stefan Kiebel
Functional Imaging Laboratory
Wellcome Dept. of Cognitive Neurology
12 Queen Square
WC1N 3BG London, UK
Tel.: +44-(0)20-7833-7478
FAX : -7813-1420
email: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|