Dear Mike,
Let me describe it in details.
The BOLD acquisition matrix is 64x64x37 (voxel sizes of 3.5x3.5x3.9). During a group analysis, spm8 produced a mask.hdr/img file, whose dimension is 79x95x86 (voxel sizes of 2x2x2) (The volume for the entire cube is 79x95x86x2x2x2=5163440; The volume for the mask is 1668672 obtained from get_totals, which is about 33% of the cube volume). To get an extent threshold to threshold this group analysis, with an alpha level of 0.05 at uncorrected p=0.001, two commands are performed:
3dClustSim -mask mask.hdr -fwhmxyz 11.2 12.0 11.1 -iter 10000 -pthr 0.001 -athr 0.05
(—> required number of voxels is 168)
3dClustSim -nxyz 79 95 86 -d 2 2 2 -fwhmxyz 11.2 12.0 11.1 -iter 10000 -pthr 0.001 -athr 0.05
(—> required number of voxels is 221)
Like you’ve said, the cube estimate is more conservative than mask estimate. Moreover, I have additional 2 questions:
(1). It seems that the extent threshold in my analysis is larger than the one reported in most fMRI research, which is usually less than 100 voxels for an alpha level of 0.05 at uncorrected p=0.001. So am I “abnormal”? Is it because the smoothness (i.e., 11.2 12.0 11.1) in my study is too large? Any method to reduce my extent threshold?
(2). AFNI 3dClustsim seems declare a new -acf argument, which requires 3 numbers. How to get the 3 numbers?
Much appreciated.
|