Hi Katharina,
> I have a linear contrast with 4 regressors modelled als A>B>C>D (3 1 -1 -3).
> Now I want to do post hoc tests only on the significant voxels in the first
> contrast, corrected for a region of interest (amygdala), for the pairwise
> contrasts (i.e. A>D and so on).
It's not very clear to me exactly which contrast you are interested in
testing, and which one you want to use to narrow the search space, or
where the ROI should come into that. But I would have thought you
could do something like this with contrast masking. E.g. using the
first contrast to inclusively mask the second.
If you have a pre-defined anatomical mask for your amygdala ROI, then
I think you could specify that as the search volume image after
clicking Small Volume Correction, so this would apply to the
correction of the main contrast. You'd probably want to use Bonferroni
for this, since I expect the voxels which are both in the ROI and
significant for the first contrast would be fairly few, and perhaps
not a smooth enough search region for the RFT maths to be valid. But
I'm guessing a little there...
Does that sound like what you want? Or have I totally misunderstood
you? Please re-post to the list (rather than directly to me) if it's
the latter.
> Therefore I would need a mask with only the
> significant T-values or with ROI-corresponding p-values, or the exact
> cutoff T-value for this ROI.
The cut-off is given as "height threshold T = ..." near the bottom of
the results table; or you can save an image of significant voxels, and
then binarise this with imcalc if you want. But again, perhaps I'm
misunderstanding what you want?
Best,
Ged
|