Hi

and apologies for posting this question again, but I really would like to know whether I am using the new option -acf correctly, both for 3dFWHMx and 3dClustSim. I am using SPM for preprocessing and statistical analyses.

For whole-brain analyses I did the following:
1. I estimated the smoothness using 3dFWHMx:
"3dFWHMx -input ResMS.nii -mask mask.nii -detrend -acf"
whereby, ResMS.nii is the group ResMS image (I did not compute the acf parameters for every subject and then compute the
average. The reason, is that I have different group-level analyses and the smoothness changes based on the
regressors I enter into the second level analysis)
mask.nii is the mask from the group-level analysis

2. I entered the estimated acf parameters into 3dClustSim:
"3dClustSim -mask mask.nii -acf 0.403842 5.06071 13.2778 -pthr 0.01 -athr 0.05 -iter 10000 -nodec -quiet"
mask.nii is the mask from the group-level analysis (same as above)

With the "classical" version (i.e. instead of using -acf I was using -fwhmxyz and entering the smoothness of the data based on the group level analysis (in mm)) I was getting a cluster size of 734 voxels for a threshold of 0.01, with -acf I get for the same group analysis and threshold a cluster of 1989 voxels.

On the other hand estimating the cluster size for an ROI with 5000 voxels, reduces the minimal cluster size from 102 voxels (classical version) to 55 voxels (-acf).

As these are significant increases/decreases in cluster size, I would like to be sure that I am estimating the cluster size correctly.

Thanks for your help and advice!
Sandra