Oh I see now. Sorry for omitting those - like you discussed at this post I was doing some testing of what confounds made sense to include so those were set in a bash script.
But, I see now that the FD was high and using the scrub strategy – which you didn’t include – was eliminating all frames. I didn’t think of it because framewise_displacement isn’t included in the confounds matrix.
I will add a check for whether confounds matrix is complete before running the model.
Thanks for your time! and happy new year