we are trying to use the Clustsim option in 3dttest++ to correct our second level model results from a flexible factorial model including 3 task conditions, 2 groups, 2 time points. As we are completely new to AFNI we are wondering about a very basic question at this point: which data format is 3dttest++ working on? We understand it says beta dataset for each individual subject, but which ones do we need here as Clustsim gives us slightly different results when looking at the three different beta images (for the three conditions) separately. But the final analyses where we’re looking to obtain a corrected threshold for are aiming at the question whether there are group differences in the T2-T1 development in brain activity.
second level model results from a flexible factorial model including 3 task conditions, 2 groups, 2 time points
Do you have three factors, one between-subjects (2 groups) and two within-subject (3 tasks and 2 time points)?
which data format is 3dttest++ working on?
Almost all AFNI programs including 3dtttest++ can take NIfTI (and AFNI) file format as input.
the final analyses where we’re looking to obtain a corrected threshold for are aiming at the question whether
there are group differences in the T2-T1 development in brain activity.
Do you mean that you’re only interested in the interaction between groups and time points (with all the 3 tasks collapsed)? If so, you can simply use T2-T1 from each subject as input for 3dttest++.
I was specifically wondering about whether I have to feed in the beta images or the contrast images. And if it’s the beta images - which ones, as the correction that we are looking for will be for the second level analyses which are based on the contrast of two betas. Would that mean I’d do the [beta for condition 1] minus [beta for condition 2] and feed this difference beta into 3dttest++?
I was specifically wondering about whether I have to feed in the beta images or the contrast images.
Depending on the context, you can feed any effect estimates as input for group analysis. It’s just a personal preference as to whether you label an effect estimate as beta value, a contrast, or a linear combination of two or more regression coefficients; you can even argue that a so-called beta value is only a contrast relative to the baseline.
the correction that we are looking for will be for the second level analyses which are based on the contrast
of two betas. Would that mean I’d do the [beta for condition 1] minus [beta for condition 2] and feed this
difference beta into 3dttest++?
Based on the information you’ve provided so far, it does sound reasonable in your case to use the contrast between those two conditions as input for 3dttest++.
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.