Hi guys,

I've used afni_proc.py to process gas challenge fMRI data. The original 3dDeconvolve model includes 6 motion derivative regressors (no gas stimulus regressor) and the errts output is then fed into an in-house software for further analyses. This software prefers time series not to be centered around 0. I've used -polort -1 to not detrend, and thus demean the time series, with detrending done in the other software. However, I'd like to try detrending in AFNI so I've included -polort 3 with the -nodmbase option but it doesn't seem to do the trick (if I understood the trick, that is) as the output time series are centered around 0. I've tried it with -stim_base in case -ortvec is for some reason not compatible as well as with non-demeaned motion regressors but same results. Where am I going wrong here?

Rick mentioned that "When using 3dD's constant polort term(s?) for the baseline, demeaning the motion parameters is important." And Bob added "For this to be the correct baseline, it is important that when you ran 3dDeconvolve to setup the matrix, you did NOT use the -nodmbase option. If you DID do that, extra baseline regressors (such as motion parameters) that you added to the model would also have nonzero means, and the baseline constant would be shared between the betas for those regressors -- making a mess." Is this consideration of the baseline important for the %SC Q here (and in context of a stimulus regressor) or does it apply to my sparse model, too?

Thank you for your help in advance!

Nic

### The original model
3dDeconvolve -input pb03.$subj.r*.scale+tlrc.HEAD \
     -ortvec mot_deriv.r01.1D mot_deriv_r01 \
     -polort -1 \
     -num_stimts 0 \
     -fout -tout -x1D X.xmat.1D -xjpeg X.jpg \
     -x1D_uncensored X.nocensor.xmat.1D \
     -fitts fitts.$subj \
     -errts errts.${subj} \
     -bucket stats.$subj

### The -nodmbase model (processing includes spatial smoothing)
3dDeconvolve -input pb04.$subj.r*.scale+orig.HEAD \
	-ortvec mot_deriv.r01.1D mot_deriv_r01 \
	-polort 3 -nodmbase \
	-num_stimts 0 \
	-fout -tout -x1D X.xmat.1D -xjpeg X.jpg \
	-x1D_uncensored X.nocensor.xmat.1D \
	-errts errts.${subj}.nodmbase \
	-bucket stats.$subj.nodmbase

### Above using the -stim_file options instead of ortvec
3dDeconvolve -input pb04.$subj.r*.scale+orig.HEAD \
	-num_stimts 6 \
	-stim_file 1 'mot_deriv.r01.1D[0]' -stim_base 1 -stim_label 1 roll 	\
	-stim_file 2 'mot_deriv.r01.1D[1]' -stim_base 2 -stim_label 2 pitch \
	-stim_file 3 'mot_deriv.r01.1D[2]' -stim_base 3 -stim_label 3 yaw 	\
	-stim_file 4 'mot_deriv.r01.1D[3]' -stim_base 4 -stim_label 4 dS 	\
	-stim_file 5 'mot_deriv.r01.1D[4]' -stim_base 5 -stim_label 5 dL 	\
	-stim_file 6 'mot_deriv.r01.1D[5]' -stim_base 6 -stim_label 6 dP 	\
	-polort 3 -nodmbase \
	-x1D X.xmat.1D -xjpeg X.jpg \
	-x1D_uncensored X.nocensor.xmat.1D \
	-errts errts.${subj}.nodmbase \
	-bucket stats.$subj.nodmbase

### Using non-demeaned motion regressors
3dDeconvolve -input pb04.$subj.r*.scale+orig.HEAD \
	-ortvec dfile_rall.1D mot_orig_withmean_r01 \
	-polort 3 -nodmbase \
	-num_stimts 0 \
	-fout -tout -x1D X.xmat.1D -xjpeg X.jpg \
	-x1D_uncensored X.nocensor.xmat.1D \
	-errts errts.${subj}.nodmbase \
	-bucket stats.${subj}.nodmbase

afni -ver: Precompiled binary linux_openmp_64: Dec 7 2020 (Version AFNI_20.3.03 'Vespasian')