Howdy-

No worries about the "pos" thing---I have done the same maaaany atime.

Re. thresholding with the F-stat: the issue here is that it is only >=0, so in converting a p-to-stat for thresholding, should one use 1sided or 2sided? (There is no issue with "bisided" here, because 2sided and bisided are exactly the same for this uniformly positive dataset, unlike if we had a t-stat or correlation or something.)

Note that there is a difference between the conversion, whether you choose 1sided or 2sided:

$ p2dsetstat -pval 0.001 -inset func_slim+orig."[0]" -1sided -quiet
6.3053
$ p2dsetstat -pval 0.001 -inset func_slim+orig."[0]" -2sided -quiet
7.01992

Note that when the GUI converts from p-to-stat, it is always using 2sidedness (again, from above, for p-to-stat conversions, 2sided and bisided are the exact same; the difference only comes in with clumping, afterwards). Testing this with func_slim*'s F-stat, when I select the following in the GUI, Thr->"Set p=0.001", the threshold goes to 7.01992. So, indeed, the GUI is hardwired to always use 2sided conversions.
If I then go through Clusterize->Rpt->SaveMsk, I do see from the command line dump that "-bisided" has been used, so the GUI is doing things consistently---it is just *always* using 2sided (or bisided), even for an F-stat.
If I instead in the GUI I use Thr->"Set p-value"->0.002, then the threshold goes to 6.3053. So, from above, we see that this "trick" of doubling my p-value P to get the equivalent of a 1sided conversion of P is needed in the GUI. That is the way you would get the "equivalent" of your first 3dClusterize command (the explicitly 1sided one) in the GUI.

Hopefully that makes sense and seems consistent. To me, the command line is clearest. One does just have to know that the GUI's p-to-stat conversions (at present) are always 2sided, and then work around that for visualization.
And indeed, I think a 1sided test on the F-stat makes sense, yes.

Also, note that you can images of all these clustering things automagically, too, with @chauffeur_afni. Some examples are here:
make [afni.nimh.nih.gov]
... even using the translucent thresholding, which appeals a lot.

--pt