Dear Stephanie,
You can do exactly thing - use A as the input image and B as a voxelwise EV. The contrasts work exactly as you have said. The only other thing worth mentioning is to be careful about the mean value. I would recommend using the -D option in randomise unless the mean value is of some interest (in which case you should model it).
All the best,
Mark
On 5 Nov 2013, at 08:18, stéphanie giezendanner <[log in to unmask]> wrote:
> Hi,
> I would like to test for a relationship between an image type A (the dependent variable) and another image type B (the concomitant variable) in a voxel-wise manner across subjects. This means my dependent variable A is a 4D image and my concomitant variable B is as well a 4D image. Is it possible to include variable B as a voxel-dependent EV in the GLM and test its relationship to the dependent variable A across subjects with randomise ? As a contrast I would set a 1 and a -1 in the voxel-dependnt EV of interest to test for a positive or a negative relationship.
> However, I am not sure whether this is correct and I havn't seen anyone implementing this so far
> Therefore, I'm looking foreward to some comments.
> Best regards,stéphanie
|