Good.
Just be aware that both the FSL results *and* the MATLAB
results are invalid while you've got NaNs like this, as in one
case they are treated like zeros and in the other case they are
treated like ones, and neither is right. To properly ignore them
you need to decrease the number of voxels being considered
in the mask, which neither method is actually doing (since they
end up with the same number of voxels in the mask).
All the best,
Mark
On 21 Sep 2009, at 03:32, Jason Stein wrote:
> Ah, that makes sense. Thanks so much for your help on that and now
> we will
> have to explore why we have NaNs.
>
> Thanks again,
> Jason
>
|