Sure, you should be able to calculate ttests, for example using 3dTcorr1D or 3dttest++, depending on how you find it easiest to compare groups/model correlation.
After that, you should be able to use 3dClustSim to estimate appropriate cluster sizes of a given significance. In terms of estimating the smoothness of noise (its FWHM), please see this previous thread/replies:
[
afni.nimh.nih.gov]
--pt