I have just used 3dmaskave to get some average response measures for voxels from a particular ROI, and the number of voxels that I am told made up the average is much larger (like 5X) than it should be (based on my manually manually counting them when I look at the mask in afni).
The input dataset and the mask are both +orig files. However, the mask was created by fractionalizing (if that's a word) an ROI originally drawn on a +tlrc dataset.... does this have something to do with the inflated voxel counts? If not, what is going on here?