On Thu, 12 Mar 2009 18:22:26 +0000, <Sarvy> <Nouri>
<[log in to unmask]> wrote:
>I have performed a two sample t-test on smoothed and modulated gray
>maps in subjects with stroke. To generate these gray matter maps, I
>outlined the stroke by hand, then used the outline as a mask during unified
>segmentation (SPM5), which resulted in the stroke mask voxels having a
>of zero in the gray matter map.
>Because the t-test ignores any voxel where even one subject has a value of
>zero, across 50 subjects, half the brain ends up being ignored because at
>least one stroke patient has injury in most brain areas.
>Is there a method that I can implement in SPM5 so that the t-test would
>include all the brain voxels rather than only a small subset (i.e. rather
>than only those voxels that are common to all the maps)?
I think the type of masking SPM does that removes voxels that have a value
of zero is called "implicit masking." If that's right, perhaps you could turn that
On the other hand, there are very good reasons not to do analysis on a voxel
if any subject doesn't have data there, so be careful.
>Any suggestions would be much appreciated