In one of our experiments we’re interested in modeling slicewise regressors along with volumewise regressors. We can get Beta coefficients for this model using 3dTfitter, but there is no stats output.

I know how to get the t-stats using Matlab, but I was wondering: Would anyone recommend a simple AFNI command for calculating t-stats after running 3dTfitter and getting Betas and model error?

Building on that (when using the default L2/least squares fitting) would it be possible to include a stats output for 3dTfitter?

Note that 3dREMLfit can do this, which might be more suitable. Or is there something special you want that only 3dTfitter can produce?

3dREMLfit has the -slibase option, for which regressors might be considered as in “type major ordering”, say, where regressors of one type are all together (one per slice), then the next set of regressors is together, again one per slice.

So if there are 5 regressors types for each of 33 slices, the first 33 regressors would be for that first slice, the next 33 for the following type, and so on.

Alternatively, and the way we use it with afni_proc.py’s ricor block to do slice-based physiological regression, is to use -slibase_sm, where ‘sm’ refers to being in slice-major order. With that option, there would be 5 regressors for slice 0, then 5 for slice 1, then 5 for slice 2, etc.

These do not make a difference in the results, but merely in how regressors are ordered in the 1D file.

Anyway, might 3dREMLfit work for you?

rick

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.