Hi,

I have being trying to make 3dLSS work for the past couple of days with no success. In Mumford (2012) paper, they suggested using a Least Square Separate approach, in which you estimate beta separately for each trial of interest with all the other trials as the other regressor.

I have an experiment in which 32 experimental conditions were presented twice in one run and 6 runs in total. Therefore, I have 32 timing flies with 6 rows and 2 time point in each row for each of the 32 conditions. I used afni_proc.py to perform the 3dDeconvolve by including 32 timing files (2 time points in each of the 6 runs). Because it is a fast event design, I want to further use 3dLSS to estimate a better beta for each of the 32 experimental conditions. I suppose I should run 32 times 3dDeconvolve followed by 3dLSS with one of the 32 conditions included as -stim_times_IM stimulus each time. OR should I just do 3dDeconvolve once with one big -stim_times_IM, in which the timing file is 6 rows with 64 time points in each row? How can I use LSS.1D to get the better beta values for each experimental condition?

The code I was using is as following.

3dDeconvolve -input pb04.$subj.r*.scale+orig.HEAD \ # 6 runs of data
-censor censor_$subj_combined_2.1D \
-polort A -float \
-local_times \
-num_stimts 2 \
-stim_times_IM 1 condition1.txt 'BLOCK(1, 1)' \ # condition1.txt contains 6 rows with two time points in each row
-stim_label 1 condition1 \
-fout -tout -x1D X.xmat.1D -xjpeg X.jpg \
-x1D_uncensored X.nocensor.xmat.1D \
-fitts fitts.$subj \
-errts errts.$subj \
-cbucket all_betas.$subj \
-bucket stats.$subj

3dLSS -matrix X.xmat.1D -save1D X.LSS.1D

Thank you so much!

Best,

Zhengang