See below.

Best Regards, 
Donald McLaren, PhD


On Wed, Oct 7, 2015 at 12:12 PM, Francesco Puccettone <[log in to unmask]> wrote:

Dear SPM list,

I have a question about the reversibility of contrasts in an analysis. It makes sense that in most fMRI studies, for every A>B contrast, the "opposite" B>A contrast is usually also checked; this is, from what I understand, due to the voxel-wise nature of the stats, and to the necessary cross-voxel averaging, which makes the differences between two conditions not "symmetrical" and thus necessitating investigation from both "ends".

The reason for investigating both sides is that SPM only displays positive values. Additionally, a voxel that is not significant for A>B could be significant for B>A. Thus, both sides need to be tested either with explicit contrasts or via viewing the negative values. The relationship of the images for A>B and B>A is the following: (A>B)*-1=(B>A).
 

 

On the other hand, sometimes, for an A>B contrast, not only the positive t-values but also the negative ones are reported, with the implication that, if B is taken as a baseline condition, the brain areas produced by this contrast are "deactivated" – a term whose intuitive interpretation there seems to be disagreement on.


Negative values for A>B means that B>A. The terms activation and deactivation are misnomers. The brain increases activity for some conditions relative others and it decreases activity to other conditions from the baseline. Changing the baseline will change the relationship with baseline, but the difference between 2 conditions doesn't change when you change the baseline.
 

 

What I am wondering is whether doing the opposite contrast is the same as looking at both negative&positive t-values in the original contrast. In other words, are negative t-values for contrast B>A the same as positive t-values for contrast A>B?

-> If so, then why are some studies doing both?

-> If not, then what exactly is the difference between them, and which one is the correct one to use when examining both "sides" of a contrast?


You have to look at both sides of your t-test. You need to know where in the brain A>B and where in the brain B>A.

 


Thanks for any help with this!


--Francesco