Hello everyone. I am having trouble understanding the statistical effects of a covariate versus a nuisance factor in freesurfer and I would appreciate some advice.

I want to run a correlation between cortical volume and the measurement of an enzyme while covarying for ETICV.

In qdec, I select volume as the measure, set the enzyme as a covariate, and select ETICV as a nuisance factor.

This works fine, however, I want to also control for the difference in using two separate scanners.

This would be a categorical variable and I understand that Freesurfer does not allow categorical variables to be nuisance variables.

The solution I found to this problem from two separate emails is that you either use the Manuel stream of mri_glmfit or use set scanner as a fixed factor, which produces a description of:

“Does the correlation between volume and enzyme, accounting for scanner, differ from zero? Nuisance factors: ETicV”

This sounds right to me, but I am confused as to the difference between “accounting for scanner” and “Nuisance factors: ETicV”, when I thought they were the same thing.

Am I approaching this right? Do I need to use mri_glmfit to properly control for scanner? If so, then how would I specify a nuisance variable in a contrast?