Hi AFNI team,
In the afni result stats.$subj* files, we can always see the q and p values and set them as the threshold.
My question is: what’s the q value fdr procedure used here? Is it Benjamini-Hochberg procedure or is it something else? Need the information for the paper method section.
And note that the default processing by afni_proc.py does not mask at the single subject level (aside from the extents mask), so those FDR values will not be so accurate. For more accuracy, run 3dFDR with a brain mask.
The output of “3dFDR -help” has some explanation. The method is the basic B-H technique (voxel-wise, not cluster-wise), with an adjustment for estimated fraction of “true negatives”. As Rick says, a mask is important to cast out non-brain voxels from the considerations.
Thank you, this answered my questions!