See inline responses below.
On Fri, Sep 14, 2012 at 4:54 AM, Julian Wiemer
<[log in to unmask]> wrote:
> Dear SPM experts,
>
> I have a problem with the analysis of a 'double contrast' like (A>B) >
> (C>D). I did this on second level with a paired t-test after creating the
> difference maps in the contrast manager.
>
> Now I want to do two things with this double contrast, but I couldn't find
> any solution for it. First, I would like to create a correlation map with a
> behavioural covariate. I know that in a paired t-test, there is the option
> to include a covariate, but you need to input 2 values per subject. If I do
> this - how do I interprete the resulting correlation map for that covariate?
> It is the activity ACROSS (A>B) and (C>D) that correlates with the
> covariate, right? But what I am looking for is the correlation of the double
> contrast with a difference between the 2 values of the covariate. Is there
> any way to do that? Can I somehow create a double contrast in a simple
> t-test?
There are two solutions:
(a) use imcalc to subtract the two contrast images for each subject;
(b) use the contrast manager for each first level model and form the
subtraction there. For example. If your contrasts were: [1 -1 0 0] and
[0 0 1 -1]. The double subtraction in the first level contrast would
be [1 -1 -1 1]. This is formed by subtracting the two contrast
vectors.
Then use a one sample t-test with a covariate. The covariate will be
the difference in the two measures you mentioned.
> And second, I would like to compare the double contrast with the same double
> contrast of another group.
Flexible factorial model in SPM or GLM Flex (from Aaron Schultz). I
generally prefer GLM Flex because its scriptable and does
between-subject effects as well, something the flexible factorial
model can't do in SPM.
>
> I hope, someone can help me, I'm really stuck with this problem. I'd
> appreciate any answers. Thank you!
>
> Julian
|