Hi,
I would like to use pre-thershold masking to constrain the number of voxels used to test for associations between brain activation and a behavioral covariate.
The mask I would like to use is the areas of significant activation from a higher-level contrast (which was cluster corrected at z 3.1, p .05). I then want to test whether, for the same contrast, there are areas within the mask that show associations with a behavioral covariate (also at the higher-level).
So far I have run my contrast, with 1 EV = group mean. Does it make sense to then run my analysis again, selecting the thresh_zstat.nii image from the previous analysis as my pre=threshold mask, with 1 EV = behavioral covariate?
For this second analysis, should the behavioral covariate be demeaned? And/or should I also include a second EV = group mean?
Also, what is an acceptable thresholding technique for this second analysis? Does it make sense to use the same thresholding as for my first analysis (cluster corrected z 3.1, p .05), or might this be too stringent? I've seen papers where authors have stated that they used a "small volume correction" for pre-threshold masking. What does this mean?
Thanks,
Sarah.
|