i have read about the importance of implementing noise smoothness information within cluster-based correction methods (particularly when using 3DClustSim). I have been running a VBM analysis and am trying to calculate the appropriate cluster size under which my findings remain valid using an initial cluster defining p<0.001 and alpha of 0.05.
My questions are:
- is the acf smoothness calculation needed for VBM studies too (as it usually refer to fMRI findings)
- could anyone provide me with help on how to implement this?
what i currently did is using the following command:
./3dClustSim -mask gm_multiple_r_mask.nii.gz -fwhm 8 -pthr .1 .05 .02 .01 0.005 0.001 -prefix 3dclustsim_gm_multiple_r_mask
as a mask i used my DARTEL template that was constructed for the present adolescent sample in order to reduce the correction to gray matter volume.
Thanks a ton on any input/help!
Can you obtain the residuals (not mean squared sum of residuals) from your VBM analysis? If so, you can use 3dFWHMx to estimate the ACF values based on the residuals from the VBM analysis, and then feed them into 3dClustSim.
Dear Gang Chen
many thanks for your feedback! I had lost my own message in the board list and only saw it now (-…-)
i read into 3DFWHMx and found some suggestions you made to a previous post, indicating that one could simply feed each subject’s file (after all the pre-processing step) for GLM as input for 3dFWHMx with option -detrend. Does that mean that I can simply replace the - input from the example with the -detrend?
set zork = (
3dFWHMx -automask -detrend junque+orig )
not sure though, what the junque+orig is referring too.
The suggestions you’re referring to in the 3dFWHMx help are for the situation when you use the residuals from the individual subject analysis as input to estimate the ACF parameters. For your case, you may try something like this with your group analysis residuals (e.g., filename group_residuals) as input:
3dFWHMx -mask yourMask -ACF group_residuals
I would be hesitant to recommend the use of 3dFWHMx on non-FMRI data – we have no experience with that, and trying to “hack” around the statistical impositions of SPM (say) by side-tracking through AFNI-land is kind of oogy IMHO.