Dear Vera,
I'm not exactly sure what the problem is here.
If it is the case that your mask (FA_GroupComparison_mask) is a subset of the full skeleton (mean_FA_skeletonised_mask) then I would expect to see less significance in the results when using the full mask (mean_FA_skeletonised_mask) as the multiple comparison correction has more voxels to account for and so the correction will be more stringent.
If this does not explain things, or my assumption about your mask (FA_GroupComparison_mask) is wrong then let us know.
All the best,
Mark
On 30 Apr 2014, at 15:53, Vera Steinheuser <[log in to unmask]> wrote:
> Dear fsl-users,
>
> I tested for an association between neuropsychological tests and FA within a group of 100 people.
> For preceding group comparisons I divided them in 2 groups. From the results of the comparisons I created a mask and used it within randomise/for the correlation analyses. This was the command:
>
> randomise i all_FA_skeletonised.nii.gz -o TestXY-FA_Correlation.nii.gz -m FA_GroupComparison_mask.nii.gz -d [Test1].mat -t [Test1].con -n 500 -c 3 -C 3 --T2 -V –D
>
> Logically, the mask of group comparisons restricted the correlations to certain structures and other were completely neglected.
>
> In order to include the structures that did not differ between the groups, we decided to use randomise again with the mean_FA_skeletonised_mask. Command:
>
> randomise i all_FA_skeletonised.nii.gz -o TestXY-FA_Correlation.nii.gz -m mean_FA_skeletonised_mask.nii.gz -d [Test1].mat -t [Test1].con -n 500 -c 3 -C 3 --T2 -V –D
>
> I am now wondering if I did it correctly since for some tests it seems that LESS voxels pop up when using the mean_FA_skeletonised_mask, when it should be the other way round (or did I get it wrong?)!
>
> Can someone let me now what the problem is and how I can fix it?
>
> Thanx,
>
> Vera
|