I would like to perform a regression to find the signal amplitudes (betas) of the task model components while at the same time removing the nuisance model components with 3dREMLfit. In order to do that, how do I have to enter the onsets of the task?
The regression matrix should be first put together by 3dDeconvolve, and then passed to 3dREMLfit.
When running an analysis with afni_proc.py, this can be done using -regress_reml_exec (and probably -regress_3dD_stop).
To do it directly with 3dDeconvolve and 3dREMLfit, just run a normal 3dDeconvolve command, but add -x1D_stop.
It will output a 3dREMLfit script that can then be run.