ETAC in 3dttest++ with covariates

AFNI Message Board
1

Can the ETAC option in 3dttest++ be used with covariates? I am interested in looking at the correlation between a symptom measure and BOLD signal in a task and have been using Clustsim, but would like to supplement the analyses with the new ETAC method.

Thank you!

2

Hi AFNI staff.

This might’ve gotten lost in the shuffle around Thanksgiving. Any advice on using ETAC with covariate measures would be much appreciated.

Thanks!

3

Yes, it can. I do not believe it cares about such things, and simply runs the permutation tests from the residuals of the regression/ttest.

  • rick
4

I’m actually trying to runs something similar. However the two etac masks did not change if I run the t-test with or without covariates. Does this mean that covariates did not affect the t-test? However, what I expected is a mask for the covariates bricks which I did not have.

5

I wanted to follow up on this thread. I ran a 3dttest++ with ETAC and the ETAC masks reflect the activity in the subbrick corresponding to the t-statistic map for the mean effect, not the t-statistic map for the covariate. I want to apply ETAC to the map that shows me the voxels where there is a relationship between task activity and a clinical correlate (i.e., voxels where there is a relationship between alcohol cue reactivity and subjects' AUDIT scores). Can someone please tell us how to get ETAC to apply to the covariate map?

6

This is a good question, and I am not even sure it is possible.

From what I know, ETAC generates random clusters from tests on permuted volumes from the residuals of a given test. Importantly, that would suggest there is only a single ETAC result for the test that includes both the main effect and the covariates. ETAC would apply to both at once, effectively, where the resulting mask might be a little different than one generated from the same test without any covariates.

If this is the case, I do not see how it would be possible to have a separate ETAC result to apply to just covariates. Though perhaps there is an aspect that I am missing...

  • rick