Hi, Brady-
Re. 1) You can run:
3dinfo -ad3 DATASET
... to see the voxelsize. Also, you might check:
3dinfo -n4 DATASET
... to see the dimensions (3 spatial, 1 temporal). Perhaps just:
3dinfo DATASET
would be useful, to view lots of the header info/properties.
Re. 4) That takeaway sounds reasonable, as a general rule of thumb. I would always try to look at the output mask and see that it is sensical: overlay it on your reference template, and see that it has good coverage, esp. in regions you care about for your analysiss.
Flipping through the 18 individual masks that comprise it would also be useful---those are shown in the APQC HTML output: in va2t if using a template (which you likely are); in ve2a if only using subject anatomical as the final "space"; in vorig if not even using a subject anatomical (rare bird). See if anyone pops out at a weird outlier, or if parts of the brain you focus on in your analysis don't get included in it (e.g., due to signal dropout---looking at *you*, subcortical nuclei!). If you have a glob over all the masks you are using, you could just use that same glob to open them in the AFNI GUI and flip through them, too.
In the APQC HTML, you can check pretty quickly on each subject's EPI-anatomical alignment ("ve2a" block) and anatomical-template alignment ("va2t" block). I would definitely check those out.
Re. 5) In most of MRI, hypotheses are of the style "is X nonzero here?" or "is the response to stimulus A different than to stimulus B here"? Those each require 2-sided testing to address: Either checking if X>0 or X<0, or if A>B or A<B, respectively. Some software do each of those subcomparisons separately, outputting two separate volumes. You wouldn't be able to use a single "bisided" clustering approach on that output directly---you would have to merge them together and do so, or (more dangerously), test each one separately *with adjusted values from what you might expect* because of the separate testing. Grumble.
If you are using a stats program in AFNI, you should generally be getting a single volume out, and you could check it for either "side" of comparison separately, but why? Using "bisided" is probably the most appropriate thing (just different than 2sided because in the former, positive and negative clumps are considered separately, even if they share boundaries). In such a case, you asked a 2-sided question, and used a 2-sided approach to answer it---and there should be no need for conversion (which just comes about with separating the question into pieces---grumble again).
Re. 6) Cool, glad that is useful.
--pt
