Hello,

I have a question about getting SD with 3dDeconvolve command. I'm using "-sresp" to get SD value, but the resulting values seem unbelievably small, whereas time-series data point from "-iresp" seem to be in the normal range.

I'd like to make sure with the experts that "-sresp" returns the standard deviation, not the standard error of the mean... I'm copying the script I ran below...

3dDeconvolve -input all_Runs+orig -xout \
-concat ../contrasts/runs.txt -polort 3 -num_stimts 16 \
-stim_file 1 ../regressors/ss4pro_regress.txt \
-stim_label 1 ss4pro \
-stim_file 2 ../regressors/ss4anti_regress.txt \
-stim_label 2 ss4anti \
-stim_file 3 ../regressors/ss4rotate_regress.txt \
-stim_label 3 ss4rotate \
-stim_file 4 ../regressors/ss8pro_regress.txt \
-stim_label 4 ss8pro \
-stim_file 5 ../regressors/ss8anti_regress.txt \
-stim_label 5 ss8anti \
-stim_file 6 ../regressors/ss8rotate_regress.txt \
-stim_label 6 ss8rotate \
-stim_file 7 ../regressors/ss12pro_regress.txt \
-stim_label 7 ss12pro \
-stim_file 8 ../regressors/ss12anti_regress.txt \
-stim_label 8 ss12anti \
-stim_file 9 ../regressors/ss12rotate_regress.txt \
-stim_label 9 ss12rotate \
-stim_file 10 ../regressors/error_regress.txt \
-stim_label 10 error \
-stim_minlag 1 0 -stim_maxlag 1 9 \
-stim_minlag 2 0 -stim_maxlag 2 9 \
-stim_minlag 3 0 -stim_maxlag 3 9 \
-stim_minlag 4 0 -stim_maxlag 4 9 \
-stim_minlag 5 0 -stim_maxlag 5 9 \
-stim_minlag 6 0 -stim_maxlag 6 9 \
-stim_minlag 7 0 -stim_maxlag 7 9 \
-stim_minlag 8 0 -stim_maxlag 8 9 \
-stim_minlag 9 0 -stim_maxlag 9 9 \
-stim_file 11 'moveParam_all.txt[1]' -stim_base 11 -stim_label 11 roll \
-stim_file 12 'moveParam_all.txt[2]' -stim_base 12 -stim_label 12 pitch \
-stim_file 13 'moveParam_all.txt[3]' -stim_base 13 -stim_label 13 yaw \
-stim_file 14 'moveParam_all.txt[4]' -stim_base 14 -stim_label 14 dS \
-stim_file 15 'moveParam_all.txt[5]' -stim_base 15 -stim_label 15 dL \
-stim_file 16 'moveParam_all.txt[6]' -stim_base 16 -stim_label 16 dP \
-iresp 1 irf_ss4pro.irf -iresp 2 irf_ss4anti.irf \
-iresp 3 irf_ss4rotate.irf -iresp 4 irf_ss8pro.irf \
-iresp 5 irf_ss8anti.irf -iresp 6 irf_ss8rotate.irf \
-iresp 7 irf_ss12pro.irf -iresp 8 irf_ss12anti.irf \
-iresp 9 irf_ss12rotate.irf \
-sresp 1 ss4pro_std -sresp 2 ss4anti_std -sresp 3 ss4rotate_std \
-sresp 4 ss8pro_std -sresp 5 ss8anti_std -sresp 6 ss8rotate_std \
-sresp 7 ss12pro_std -sresp 8 ss12anti_std -sresp 9 ss12rotate_std \
-fitts full_model.fit -errts residual_error.fit \
-rout -fout -tout \
-bucket decon_results


I have 9 trial-types, plus error and motion parameters. Each trial type has between 20-30 trials. It is puzzling to me why we get normal-looking time series data with extremely low SD. This is my first deconvolution study (as opposed to trial-averaging). Is SD in deconvolution normally smaller than trial averaging?

Thanks so much in advance! Akiko