Hi Nan,
If I'm understanding you correctly, you are seeing differences between A masked with B>0 and B masked with A>0. This is what I would expect to see. Consider dividing the voxels up into different categories:
1 - A>threshold, and B>threshold (i.e. both significant)
2 - A>threshold, and 0<B<threshold
3 - B>threshold, and 0<A<threshold
4 - 0<A<threshold, and 0<B<threshold
5 - All others
If you test A masked with B (not B>0, but B) then you'd only see voxels in category 1.
If you test A masked with B>0 then you'd see voxels in categories 1 and 2.
If you test B masked with A>0 then you'd see voxels in categories 1 and 3.
You would never see voxels in categories 4 or 5, but the variations in your results are likely to be due to the voxels in categories 2 and 3.
All the best,
Mark
On 26 Jun 2014, at 21:57, Nan Wise <[log in to unmask]> wrote:
> Dear FSL experts,
>
> Thanks for your advice about contrast masking on the second level. I have done it with good effect and have been able to mask the contrast A>B with A>O and B>O, thereby assuring that only the voxels greater than 0 show up in the A>B contrast.
>
> One last question--regarding conjunction analyses--
>
> I am looking at two conditions A>0 (baseline) and B>0 (baseline) -- with the desire to know what areas are similarly active.
>
> On the second level, I set up 2 EVs--(1) A and (2) B and do contrasts for the group means for A and for B. Then I contrast masked (voxels >0) A with B and then B with A. I sometimes get different results for the A masked with B and the B masked with A.
>
> Can you explain how that happens and what that means? Not sure how to interpret it as I thought conjunction means only the voxels that are common to both contrasts will be shown.
>
> Thanks so much for your help.
>
> Two months to my defense and I am winding down the analyses and writing now.
>
> Much appreciation-
> Nan Wise
> Rutgers-Cognitive Neuroscience
|