adding a covariate in the 3dttest++ command : strange results


I’m emailing you cause I have strange results when adding a covariate in my 3dttest++ command.
Before adding it , results look as expected : ; but then with the covariate it looks like the field of view in being changed, (attached file) do you know what’s happening ?

Thank you in advance

Hi, Lysianne-

Hmm, interesting. What do the images look like unthresholded?


oh, I haven’t look at that before , is it because the posterior brain areas were not sufficiently activated ?

Wow, powerful covariate! And to check one further thing, the range in the colorbars is the same, I guess?

That is an oddly focal change to have with the covariate inclusion. Hmm…


Like too strong to be correct
hm the color bars are not the same actually, I just let the default set them up
the one without the covariate has a max at 2.027
the one with the covariate has a max of 3.26

though looks the same when I scaled them from [-4:4]

Is there some sort of group masking that is relevant here? The dark (higher magnitude) area looks a bit like a coverage intersection box. Are various subjects missing from coverage outside of that? Do you have a group mask? Do you have a group ‘sum’ mask (add up subjects per voxel)?

EDIT: For reference, it is nice to run something like:

3dMean -prefix QC/mask.mean sub*/*.results/mask_epi_anat*.HEAD

Which is part of the group QC we were using (a few years back) in AFNI_demos/AFNI_pamenc/AFNI_02_pamenc/global_process_outline.txt

  • rick

Thanks, that is mildly illustrative that the situation is: adding the covariate makes that swath have notably larger-magnitude values than the surroundings, rather than it shrinks values outside the swath.

I am still not sure what would cause this. Some kind of outlier-ish interaction with values? Might have to look at the data to investigate further.