Hi Megan,
There are just a few things to note here, and the main one is that 1D (TSV) inputs should be concatenated across runs.
For example, the 4 sub-1055_task-learn_run-*_motion.tsv should be concatenated, with only the 1 input passed via -regress_motion_file. I _could_ consider allowing 1 per run as an alternative, but it is not an option right now.
Also, -regress_extra_ortvec can take multiple files as input, but they should also be concatenated across runs. So if you were to :
cat sub-1055_task-learn_run-*_aCompCor6.tsv > ortvec_rall_aCompCor6.1D
(and similarly for cosine and fd), then the options might read:
-regress_extra_ortvec ortvec_rall_aCompCor6.1D \
ortvec_rall_cosine.1D \
ortvec_rall_fd.1D \
-regress_extra_ortvec_labels aCompcor6 cosine fd \
Note that the "labels" option just names the ortvec files in some way.
Also, the -regress_stim_times_offset parameter is probably negated. If fmriprep has done 3dTshift to align to the middle(ish) of the TR, then it will be as if the EPI timing is later than the start of the TR, and so relatively, the task timing should be earlier. Consider using:
-regress_stim_times_offset -0.870967741935484
Also, just to make it easier, since the -regress_stim_types and -regress_basis_multi parameters are all the same, it is okay to use a single parameter, as in:
-regress_stim_types AM1 \
-regress_basis_multi 'dmBLOCK(0)' \
Does that seem reasonable?
- rick