Hello. I use an event-related paradigm with different stimulus classes to be compared.

Using the following script:

*******
foreach subject ( cr la )
cd ../${subject}*
if ( -e gltreg+orig.BRIK ) then
rm -rf gltreg+*
endif
3dDeconvolve -input par12newdsbpf+orig -concat ../vectors/runs.1D -mask 3dCL_mask+orig -nfirst 0 -num_stimts 19 \
-polort 1 \
-stim_file 1 ../vectors/'neuant.1D' -stim_label 1 'neuant' \
-stim_file 2 ../vectors/'allrewant.1D' -stim_label 2 'allrewant' \
-stim_file 3 ../vectors/'allpunant.1D' -stim_label 3 'allpunant' \
-stim_file 4 ../vectors/'win5ant.1D' -stim_label 4 'win5ant' \
-stim_file 5 ../vectors/'win20ant.1D' -stim_label 5 'win20ant' \
-stim_file 6 ../vectors/'lose5ant.1D' -stim_label 6 'lose5ant' \
-stim_file 7 ../vectors/'lose20ant.1D' -stim_label 7 'lose20ant' \
-stim_file 8 ../vectors/'winmysant.1D' -stim_label 8 'winmysant' \
-stim_file 9 ../vectors/'losemysant.1D' -stim_label 9 'losemysant' \
-stim_file 10 'rewardhits.1D' -stim_label 10 'rewhits' \
-stim_file 11 'rewardmiss.1D' -stim_label 11 'rewmiss' \
-stim_file 12 'losshits.1D' -stim_label 12 'losshits' \
-stim_file 13 'lossmiss.1D' -stim_label 13 'lossmiss' \
-stim_file 14 '3dmotionnew.1D[1]' -stim_label 14 'roll' \
-stim_file 15 '3dmotionnew.1D[2]' -stim_label 15 'pitch' \
-stim_file 16 '3dmotionnew.1D[3]' -stim_label 16 'yaw' \
-stim_file 17 '3dmotionnew.1D[4]' -stim_label 17 'dS' \
-stim_file 18 '3dmotionnew.1D[5]' -stim_label 18 'dL' \
-stim_file 19 '3dmotionnew.1D[6]' -stim_label 19 'dP' \
-tout -bucket gltreg \
-glt 1 ../vectors/allrewvneu.mat -glt_label 1 'rewvneuant' \
-glt 1 ../vectors/allpunvneu.mat -glt_label 2 'punvneuant' \
-glt 1 ../vectors/hvlrewant.mat -glt_label 3 'hvlrewant' \
-glt 1 ../vectors/hvlpunant.mat -glt_label 4 'hvlpunant' \
-glt 1 ../vectors/mv20rewant.mat -glt_label 5 'mv20rewant' \
-glt 1 ../vectors/mvnrewant.mat -glt_label 6 'mvnrewant' \
-glt 1 ../vectors/mv20punant.mat -glt_label 7 'mv20punant' \
-glt 1 ../vectors/mvnpunant.mat -glt_label 8 'mvnpunant' \
-glt 1 ../vectors/gainfbk.mat -glt_label 9 'gainfbk' \
-glt 1 ../vectors/lossfbk.mat -glt_label 10 'lossfbk' \

******
together with single line matrices (a la 3dDeconvolve example 1.4.4.6) to detect voxels where AUC of stimulus type "rew" (stim_file 2 waveform) is greater than stimulus class "neu" (stim_file 1 waveform), e.g.

0 0 0 0 -1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.

This results in a comprehensive bucket dataset with paired "coef" and "t-st" subbricks for each stimulus class singly as well as the specificed glt contrasts. [LC] t-st activation maps show "rew"-responsive voxels compared to "neu", for example.

I would like to compare the strength/magnitude of this glt contrast between patient group A and patient group B.

3dttest is an easy and straightforward program for comparing Beta coefficients with a single idealized waveform (assuming regressors were applied to a dataset normalized as % signal change).

However, the glt gives me some kind of "net active" voxels showing preference (AUC) for "rew" over "neu".

My question is, is it mathematically valid (or "reasonable" to set up a between-group t-test on the coef of an [LC] *contrast*?

What might be a reasonable solution to detect voxels showing greater rew vs neu activation in Patient group A compared to activation in group B? It seems like an LC sub-brick "coef" is different in character from the "coef" with a single stimulus class waveform.

Jim Bjork