Reviving this as well:

1. Do you guys recommend using ANATICOR with task data? It usually reduces effect sizes for us but that does not mean that it's not closer to the truth... We implement it in afni_proc.py by:

-regress_anaticor_fast \
    -regress_reml_exec \

2. When having empty stimfiles (e.g. due to an event never occurring, like failing a certain task) you can prevent 3dDeconvolve to crash by adding by adding e.g.

-regress_opts_3dD \
		-allzero_OK \
		-GOFORIT 6 \

The anaticor regressors are implemented after 3dDeconvolve by:

# -- execute the 3dREMLfit script, written by 3dDeconvolve --
# (include ANATICOR regressors via -dsort)
tcsh -x stats.REML_cmd -dsort Local_WMe_rall+tlrc

Is there a similar way to tell this process to also ignore/allow these empty stimfiles? Because now the 3dDeconvolve part is fine but it crashes at the 3dREMLfit part:

tcsh -x stats.REML_cmd -dsort Local_WMe_rall+tlrc
3dREMLfit -matrix X.xmat.1D -input pb03.subject_pbo_06.r01.volreg+tlrc.HEAD -fout -tout -Rbuck stats.subject_pbo_06_REML -Rvar stats.subject_pbo_06_REMLvar -Rfitts fitts.subject_pbo_06_REML -Rerrts errts.subject_pbo_06_REML -verb -dsort Local_WMe_rall+tlrc
++ 3dREMLfit: AFNI version=AFNI_16.3.04 (Oct 19 2016) [64-bit]
++ Authored by: RWCox
++ Number of OpenMP threads = 12
++ No mask ==> computing for all 172800 voxels
++ FDR automask has 60405 voxels (out of 172800 = 35.0%)
** ERROR: matrix column #12 is all zero!?
** ERROR: matrix column #14 is all zero!?
** FATAL ERROR: Can't continue with all zero columns without -GOFORIT option!
** Program compile date = Oct 19 2016

Thanks!