GLM over volume of surfaces

Dear AFNI users,
I have a nifti with 200 volumes of spatial maps corresponding to different subjects, along with a design matrix of 200 rows, in the same order as the nifti file, of 10 covariates. Is there functionality for performing something like 3dttest++ using only these files? Basically, I want to perform an ordinary linear regression and only obtain t-stats and p-values for each covariate.

The linear regression programs in AFNI for such cases appear to all assume temporal dependence. As such, I’m unsure if I can use those programs.
Thank you.

Hi Chintan,

Such a linear regression can be done with either 3dttest++
or 3dDeconvolve (with demeaned covariates). Using 3dttest++,
the input datasets should be separate, as in there should be
200 datasets (see 3dTsplit4D to separate a dataset). And the
covariate file entries would need to be tied to each of those
datasets, preferably by a subject ID.

Using 3dDeconvolve, the input should be a single dataset,
and the covariates (regressors) should all have zero mean
(or whatever center is appropriate).

Since this is a group analysis, 3dttest++ would be a more
typical way to run the regression. However, it looks like
you have set it up such that 3dDeconvolve would be easier.
In your place, I might be inclined to set it up for 3dttest++.

  • rick

Thank you Rick!