Hi-
To me, #2 seems to make the most sense:
+ Step 1: calculate the RSFC parameters (fALFF, mALFF, etc.) across the whole brain *once*.
+ Step 2: assuming MASKNAME is a (small, or at least chosen) region of interest, then calculate the mean RSFC parameter of interest in that masked region.
... and repeat Step 2 for each region of interest you have.
You can calculate the mean RSFC in several, non-overlapping regions simultaneously, if you want, by making sure that each region is made up of voxels of a single integer value, and that each region uses a different integer. For example, one region of all '1' values, another region of all '2', etc. Then, 3dROIstats will calculate the mean RSFC parameter for each region separately.
If you have a single map to threshold and obtain many clusters from, you can do the above (threshold to determine regions, label each separate region with a distinct integer, even get rid of regions volumes that are 'too small) using 3dROIMaker:
[
afni.nimh.nih.gov]
But it really depends on how many regions you want to calculate parameters for in your study.
--pt