Question regarding TENTzero results

[carolin31]

Dear afni experts,

I am using multiple basis functions for my task based fMRI analysis. GAM for the regressors I don t care about, and TENTzero (0,15,7) for the regressors I do care about. When using only TENT for all regressors, I had previously some collinearity issues in the design matrix.

I am confused about the TENTzero results:
I get one Fstat and 5 Coef and a Tstat (#0 - #4) for each of the regressors, which I modeled with TENTzero. I assume that it is 5 because the first and last TENT should be 0? But, I also see another Fstats and Coef and Tstat with the number #0 for the same regressor. See number 10 and 11 and 32 and 33 in the attached image. Why are there two times coef #0 and Tstats #0 for my regressors and what is the difference?

Thank you very much
Carolin

[Gang]

I get one Fstat and 5 Coef and a Tstat (#0 - #4) for each of the regressors, which I modeled with TENTzero.
I assume that it is 5 because the first and last TENT should be 0?

Yes, with TENTzero you assume the first (stimulus onset) and last (recovery) TENT is of zero magnitude.

Why are there two times coef #0 and Tstats #0 for my regressors and what is the difference?

That’s strange indeed. Are they identical?

There are also other repetitions, for example, #20 vs. #34, #21 vs. #35, #22 vs. #36, and #31 vs. #37. We may have to see the 3dDeconvolve script to understand why that’s the case.

[carolin31]

Hello Gang,

thank you very much for your quick response.
I just noticed something! The screenshot is showing how the bucket dataset looks like after REML. If I check the dataset after 3dDeconvolve, it looks different (see attachment). There is a GLT included in all regressor names in the right column.
I am not sure why this is not displayed in the dataset after REML.

And yes, the dataset with the same names in REML is not identical.

Is there something that I need to change in my script, so that it displays the right names in the REML dataset as well?
And does it mean that for the estimation of the coefficients and Tstats from the GLT regressors, the 5 TENTs were summarized and used together?

Thank you very much again!

Below you find the 3dDeconvolve script in case you still want to look at it:

3dDeconvolve -input pb05.subj.r*.scale+orig.HEAD \ -censor censor_{subj}_combined_2.1D
-polort 2 -float
-local_times
-num_stimts 41
-stim_times 1 stimuli/stimtimes.01.1D ‘GAM’
-stim_label 1 Obj_T_Hit
-stim_times 2 stimuli/stimtimes.02.1D ‘GAM’
-stim_label 2 Obj_T_M
-stim_times 3 stimuli/stimtimes.18.1D ‘GAM’
-stim_label 3 Obj_no_response
-stim_times 4 stimuli/stimtimes.19.1D ‘TENTzero(0,15,7)’
-stim_label 4 Lure_Obj_FA
-stim_times 5 stimuli/stimtimes.20.1D ‘TENTzero(0,15,7)’
-stim_label 5 Lure_Obj_CR
-stim_file 6 mot_demean.r01.1D’[0]’ -stim_base 6 -stim_label 6 roll_01
-stim_file 7 mot_demean.r01.1D’[1]’ -stim_base 7 -stim_label 7 pitch_01
-stim_file 8 mot_demean.r01.1D’[2]’ -stim_base 8 -stim_label 8 yaw_01
-stim_file 9 mot_demean.r01.1D’[3]’ -stim_base 9 -stim_label 9 dS_01
-stim_file 10 mot_demean.r01.1D’[4]’ -stim_base 10 -stim_label 10 dL_01
-stim_file 11 mot_demean.r01.1D’[5]’ -stim_base 11 -stim_label 11 dP_01
-stim_file 12 mot_demean.r02.1D’[0]’ -stim_base 12 -stim_label 12 roll_02
-stim_file 13 mot_demean.r02.1D’[1]’ -stim_base 13 -stim_label 13
pitch_02
-stim_file 14 mot_demean.r02.1D’[2]’ -stim_base 14 -stim_label 14 yaw_02
-stim_file 15 mot_demean.r02.1D’[3]’ -stim_base 15 -stim_label 15 dS_02
-stim_file 16 mot_demean.r02.1D’[4]’ -stim_base 16 -stim_label 16 dL_02
-stim_file 17 mot_demean.r02.1D’[5]’ -stim_base 17 -stim_label 17 dP_02
-stim_file 18 mot_demean.r03.1D’[0]’ -stim_base 18 -stim_label 18 roll_03
-stim_file 19 mot_demean.r03.1D’[1]’ -stim_base 19 -stim_label 19
pitch_03
-stim_file 20 mot_demean.r03.1D’[2]’ -stim_base 20 -stim_label 20 yaw_03
-stim_file 21 mot_demean.r03.1D’[3]’ -stim_base 21 -stim_label 21 dS_03
-stim_file 22 mot_demean.r03.1D’[4]’ -stim_base 22 -stim_label 22 dL_03
-stim_file 23 mot_demean.r03.1D’[5]’ -stim_base 23 -stim_label 23 dP_03
-stim_file 24 mot_deriv.r01.1D’[0]’ -stim_base 24 -stim_label 24 roll_04
-stim_file 25 mot_deriv.r01.1D’[1]’ -stim_base 25 -stim_label 25 pitch_04
-stim_file 26 mot_deriv.r01.1D’[2]’ -stim_base 26 -stim_label 26 yaw_04
-stim_file 27 mot_deriv.r01.1D’[3]’ -stim_base 27 -stim_label 27 dS_04
-stim_file 28 mot_deriv.r01.1D’[4]’ -stim_base 28 -stim_label 28 dL_04
-stim_file 29 mot_deriv.r01.1D’[5]’ -stim_base 29 -stim_label 29 dP_04
-stim_file 30 mot_deriv.r02.1D’[0]’ -stim_base 30 -stim_label 30 roll_05
-stim_file 31 mot_deriv.r02.1D’[1]’ -stim_base 31 -stim_label 31 pitch_05
-stim_file 32 mot_deriv.r02.1D’[2]’ -stim_base 32 -stim_label 32 yaw_05
-stim_file 33 mot_deriv.r02.1D’[3]’ -stim_base 33 -stim_label 33 dS_05
-stim_file 34 mot_deriv.r02.1D’[4]’ -stim_base 34 -stim_label 34 dL_05
-stim_file 35 mot_deriv.r02.1D’[5]’ -stim_base 35 -stim_label 35 dP_05
-stim_file 36 mot_deriv.r03.1D’[0]’ -stim_base 36 -stim_label 36 roll_06
-stim_file 37 mot_deriv.r03.1D’[1]’ -stim_base 37 -stim_label 37 pitch_06
-stim_file 38 mot_deriv.r03.1D’[2]’ -stim_base 38 -stim_label 38 yaw_06
-stim_file 39 mot_deriv.r03.1D’[3]’ -stim_base 39 -stim_label 39 dS_06
-stim_file 40 mot_deriv.r03.1D’[4]’ -stim_base 40 -stim_label 40 dL_06
-stim_file 41 mot_deriv.r03.1D’[5]’ -stim_base 41 -stim_label 41 dP_06
-iresp 4 iresp_Lure_Obj_FA.$subj
-iresp 5 iresp_Lure_Obj_CR.$subj
-GOFORIT 16
-allzero_OK
-num_glt 5
-gltsym ‘SYM: +Lure_Obj_FA[1…3]’
-glt_label 1 Lure_Obj_FA
-gltsym ‘SYM: +Lure_Obj_CR[1…3]’
-glt_label 2 Lure_Obj_CR
-gltsym ‘SYM: +0.5Lure_Obj_FA[1…3] -0.5Lure_Obj_CR’
-glt_label 3 LureFA_minus_CR
-gltsym ‘SYM: +0.5Lure_Obj_CR[1…3] -0.5Lure_Obj_FA’
-glt_label 4 LureCR_minus_FA
-gltsym ‘SYM: +0.5Lure_Obj_FA[1…3] +0.5Lure_Obj_CR’
-glt_label 5 LureCR_FA
-jobs 8
-fout -tout -x1D X.xmat.1D -xjpeg X.jpg
-x1D_uncensored X.nocensor.xmat.1D
-fitts fitts.subj \ -errts errts.{subj}
-bucket stats.$subj

[Gang]

Carolin, for some reason 3dREMLfit does not attach the GLT index number in the sub-brick label as 3dDeconvolve does. The sub-bricks #32-46 correspond to the following post hoc tests you specified in 3dDeconvolve with 3 sub-bricks for each test (effect, t-stat, and F-stat):

-glt_label 1 Lure_Obj_FA \
-gltsym ‘SYM: +Lure_Obj_CR[1…3]’ \
-glt_label 2 Lure_Obj_CR \
-gltsym ‘SYM: +0.5Lure_Obj_FA[1…3] -0.5Lure_Obj_CR’ \
-glt_label 3 LureFA_minus_CR \
-gltsym ‘SYM: +0.5Lure_Obj_CR[1…3] -0.5Lure_Obj_FA’ \
-glt_label 4 LureCR_minus_FA \
-gltsym ‘SYM: +0.5Lure_Obj_FA[1…3] +0.5Lure_Obj_CR’ \
-glt_label 5 LureCR_FA \