whereami and 3dgroupincorr

AFNI Message Board
1

Hello! When I use whereami, sometimes regions are output saying they have 0% overlap with the blob I'm analyzing. How should I interpret this? If it listed all of the ROIs in the atlas and many of them were zeros, that would be clear to me. But it usually just chooses a few to list as zeros. Is this related to rounding or partial voxels perhaps?

I also have a couple questions about 3dGroupInCorr.

  1. I see that that AFNI defaults to Talairach space in this program, but original space is an option by updating an environment variable to "YES". Is there an option to set it to MNI space instead?

  2. When I ask for it to output the t-test comparison for two similar maps, the results seem off. Most of the voxels in the brain survive even very conservative thresholding, and then all of the voxels cut off at a z-score threshold of 13. Does this have something to do with the Fisher transform that's performed?

Thank you!

2

Hi! Following up to ask whether these questions can be answered, please and thank you!

3

Hello,

The overlap values of 0% should mean that less than 1/2% of overlap exists.

I believe those numbers are printed via "whereami -omask", run on a resampled atlas (resampled to match the current data). The percentages are output with 1 decimal place of accuracy. So a value of 0.0 means within the range (0, 0.05), excluding those endpoint values.

Regarding 3dGroupInCorr:

  1. When a view extension is +tlrc, that mean some standard space, possibly Talairach, but also possibly MNI or a monkey or a rat template. It is up to the user to decide which template to use for warping to standard space.
    What template did you specify in the subject level analysis?

  2. How many subjects are you using? For a t-test comparison, the seed location should have a result that is basically zero.
    13 is actually a somewhat magical number here. It is the largest z-score output by a Fisher transformation (at that point, it is viewed as going off toward infinity, as the computations do not have the precision to go so far from zero).

The fact that you are seeing lots of 13's is surprising. How many subjects are included in this test?

  • rick