I am a little confused when running afni_proc where the corr_brain file comes from. Is this automatically generated when using the anaticor option? Is this giving you a whole brain correlation value? And what is the value given by 3dinfo at sub-brik 0? it is a range and different from the GCOR value that is in the output script.
If I have already pre-processed my datasets using afni_proc.py,(and have the corr_brain datasets)… is there a way I can regress additional correlation volumes (for gray matter specifically) without having to run the entire process all over again? Could I use these datasets for a group comparison of gray matter correlation and use the values to correlate with behavioral scores?
Thanks so much in advance!
The corr_brain volume is just the average correlation between
each current voxel and every other voxel in the EPI masked
dataset. It is created anytime the full_mask and errts
The GCOR value is the average of the corr_brain dataset over
that same mask.
To be sure, where would you be regressing the correlation
volumes, would they be voxelwise covariates at the group
Or do you mean you would like to create additional correlation
I meant I would like to create additional correlation volumes - so correlation only in the gray matter, for instance. Would it then be possible to plug these datasets into 3dttest++ to look at group differences in correlation without having to do s seed-based analysis? And to create these additional volumes, would I have to run afni_proc all over again or would I just use command below substituting full_mask for a gray matter mask?
3dTnorm -norm2 -prefix rm.errts.unit errts.$subj.anaticor+orig
3dmaskave -quiet -mask full_mask.$subj+orig rm.errts.unit+orig
3dTstat -sos -prefix - gmean.errts.unit.1D' > out.gcor.1D
echo “-- GCOR =
compute correlation volume
(per voxel: average correlation across masked brain)
(now just dot product with average unit time series)
3dcalc -a rm.errts.unit+orig -b gmean.errts.unit.1D -expr ‘a*b’ -prefix rm.DP
3dTstat -sum -prefix corr_brain rm.DP+orig
The rm.errts.unit dataset could be used for any of these
correlation volumes. The one line that would really need
to be altered for each mask is the 3dmaskave command.
Beyond that, it is just keeping track of the file names.
Plugging those (possibly Fisher transformed datasets) into
3dttest++ would be akin to a seed-based analysis, where
seed ROI is the mask.