Experts, I want to calculate % signal change rate using regression coef. after 3dDeconvolve(as the way decribed by Gang Chen in 'Scaling and Percent Signal Change'). Could you tell me how to get the baseline constant? In the following two, which one, the former or the latter is right?

3dDeconvolve -input taskcalc+orig -polort 1 -progress 1000 -num_stimts 7 \
-stim_file 1 taskwaver.1D -stim_label 1 task \
-stim_file 2 motion.1D[0] -stim_label 2 roll -stim_base 2 \
-stim_file 3 motion.1D[1] -stim_label 3 pitch -stim_base 3 \
-stim_file 4 motion.1D[2] -stim_label 4 yaw -stim_base 4 \
-stim_file 5 motion.1D[3] -stim_label 5 dl -stim_base 5 \
-stim_file 6 motion.1D[4] -stim_label 6 ds -stim_base 6 \
-stim_file 7 motion.1D[5] -stim_label 7 dp -stim_base 7 \
-tout -fout -bout -bucket task.buck

or

3dDeconvolve -input taskcalc+orig -polort 1 -progress 1000 -num_stimts 1 \
-stim_file 1 taskwaver.1D -stim_label 1 task \
-tout -fout -bout -bucket task.buck

And the -input 3D-time data "taskcalc+orig" is gotten through 3dvolreg motion correction and 3dcalc signal change normalizing. Is it right?

I am confused after read
[afni.nimh.nih.gov]; (The correct baseline is 100, but you are thinking it is 14. So instead of reporting a 3 percent response for some other stimulus type, you report 21. "Betas gone wild"...)

and
"[afni.nimh.nih.gov]; (use the motion corrected time series in the 3dDeconvolve -input . The motion regressors are for the residual effects of motion.)

Thank you very much in advance.