I've read both Hagler et al 2006 Neuroimage (Smoothing and cluster thresholding for cortical surface-based group analysis of fMRI data) and B. Douglas Ward's description of AlphaSim that both describe clusterwise correction for multiple comparisons.

I recently found, however, an MGH Freesurfer presentation that mentioned that after synthesizing Gaussian noise and smoothing (Monte Carlo), the rows of the design matrix were permuted.

Does the mri_glmfit-sim clusterwise correction method permute rows of the design matrix? I'm still a bit confused about the exact method the clusterwise correction for multiple comparisons work in Freesurfer.

Thanks for your help.

Christopher Hyatt, PhD
IOL/Olin Neuropsychiatry Research Center
Hartford, CT