Hi!
If I would have normal fMRI-data and run a factorial 3dMVM analysis I know the steps to get "valid" cluster-corrections using the errts-files to in 3dFWHMx -ACF to generate the acf-paramters followed by simulations in 3dClustSim where you have to pick a "maximum" a per-voxel threshold of 0.002 to be able to claim that the results are false-pos corrected according to AFNI-guidelines.
But this time we have single 3D volume structural maps (from the FSL-VBM pipeline). We now want to do a 3x2 factorial analysis using 3dMVM. The analysis went fine but how do we tackle the cluster-correction question here?
Our approach, since this is not bold-data, is to use the data itself with brain-masks to get smoothness via 3dFWMHx -ShowMeClassicFWHM . We argued that we can use the old -FMHM method since this is not temporal bold-data. I.e. get the x-y-z smoothness estimates and plug them into 3dclustsim -fwhmxyz with the group-test-mask used in 3dMVM.
Would this be correct? If so, is there any "maximum" allowed per voxel p-value to use, like the 0.002 in the fMRI guidelines?
Best,
Robin