I'm not quite clear on what the -rankize option on 3dttest++ does. Here are my questions:

1) How exactly are they ranked? Is it such that, if you had 15 subjects, each voxel in the dataset would be transformed into a rank value (1-15), based on its value in comparison to the other subjects?
2) Is the ranking across groups? For my analysis, I have two groups, and I would prefer for both the data and the covariates to be ranked within each group.

If it helps, what I'm trying to accomplish is a moderation analysis, where I'm looking for areas in the brain where the relationship between the brain response (dependent variable) and a behavioral measure (covariate) is moderated by group. I wanted to use slopes based on a Spearman correlation due to outliers/nonnormal data.

Thank you.