3dFDR on F-statistics

Hi AFNI team,

I’m using 3dFDR to threshold my stats files for the first time, and I would like to confirm with you that I’m using it correctly. I don’t understand something happened in the results…

I used 3dDeconvolve with -xsave option to calculate a new contrast. This contrast is something like +condition1 +condition2 +\condition3 … So in the end, I care about the F stats sub-brick that shows either of these conditions significant.

The result stats file (I’ll call it statsfile) didn’t give me q values, but only p values and F stats in the sub-brick I’m interested. So I applied 3dFDR to it using the command:

3dFDR -input statsfile -mask mask_anat.file -prefix statsfile_fdr

With the result statsfile_fdr, I checked the same F-stats sub-brick. I found the values become smaller in general, which I assume it’s because previous F-values got recalculated to fdr z values. There is still no q-values, but only p-values. Here, I assume the new p-values should be q-values.

However, I found, if I threshold the new statsfile_fdr file with p=0.05 (which should be q=0.05), the result mask is actually even larger than p=0.05 with the previous statsfile. Is this even possible? Or did I understand this wrongly somewhere?

Thank you so much!