Correlation/Regression: FC map vs predictor

Dear AFNI experts,

my aim is to evaluate if a zero-centered variable (X) predicts the topography of functional connectivity (FC) maps across subjects.
More specifically: I obtained FC maps from 40 subjects from a seed region; each subject is associated with a value in X. FC maps are unthresholded z-Fisher transformed maps from r values (no zeros).
I want:
(i) to statistically test if X predicts the topography of FC maps;
(ii) to obtain predicted FC maps at extreme values of X;
(iii) if possible, to highlight specific spatial elements (voxels) which contribute to such prediction.

My guess was generating voxel-wise subject-specific maps with constant values (= values of X) and then using these as voxel-wise covariates in 3dttest++. But actually, I’m not sure it satisfies point (i), and quite sure it won’t satisfy (ii) and (iii).
Is 3dttest++ appropriate for the purpose? Does something exist which is appropriate for the purpose? Any help is appreciated.

Thanks in advance,