Hello AFNI helpers, I am in need of your assistance.

I believe I have diagnosed my problem (with guidance from other messages), but I'm having trouble with the solution. In short, I'm having difficulty with my stim files / design matrix.

One of my regressors is a linear combination of another two, but I'm not sure how to get away from this.

The experimental is as follows:
Two groups, Patient(P) and Controls(C), perform a discounting task where they make a choice between a Small-Sooner(SS) or Large-Later choice(LL), the choice is presented for variable durations (their RT for each choice).

I was hoping to model separately the trials they made :
an SS-choice
an LL-choice,
and doing amplitude modulation of every event by
1. sum(subjective values),
2. difference(subjective values), and
3. delay(LL-choice)
- with every event duration modulated by the RT.

-stim_times_AM1 1 ../stimuli/${subj}SS.1D 'dmBLOCK' -stim_label 1 SS \
-stim_times_AM1 2 ../stimuli/${subj}LL.1D 'dmBLOCK' -stim_label 2 LL \
-stim_times_AM2 3 ../stimuli/${subj}stimlist2.1D 'dmBLOCK' -stim_label 3 ALL \

I also tried putting the SS and LL files into the ALL stimfile (as 1s or 0s when they occurred), but that is the same problem - and what made me realize where the linear combination was (constant response = SS + LL).

The comparison of interest is the SS vs LL trials - to see if there is different activation between Patients and Controls, without the influence of differences in subjective value.

One option might be to do Amplitude Modulation (by sum, diff, delay) for SS and LL separately, but wouldn't I want those modulations (sum, diff, delay) to be modeled the same way throughout the entire session? I guess I'll give that a try and see if I have some intuition for what it means when I look at the design matrix. I'm sorry if I'm being dense, but this seems like something I should understand better before moving forward.

Thanks so much for any and all guidance.
Hugo Decker

Here is the full output: (there are also collinearity Warnings, but those I understand and can't be helped)


3dDeconvolve -input ../pb03.01.r01.scale+tlrc.HEAD ../pb03.01.r02.scale+tlrc.HEAD ../pb03.01.r03.scale+tlrc.HEAD ../pb03.01.r04.scale+tlrc.HEAD -censor ../motion_all_censor.1D -polort 3 -float -global_times -num_stimts 9 -stim_times_AM1 1 ../stimuli/01SS.1D dmBLOCK -stim_label 1 SS -stim_times_AM1 2 ../stimuli/01LL.1D dmBLOCK -stim_label 2 LL -stim_times_AM2 3 ../stimuli/01stimlist2.1D dmBLOCK -stim_label 3 ALL -stim_file 4 motion_demean.1D[0] -stim_base 4 -stim_label 4 roll -stim_file 5 motion_demean.1D[1] -stim_base 5 -stim_label 5 pitch -stim_file 6 motion_demean.1D[2] -stim_base 6 -stim_label 6 yaw -stim_file 7 motion_demean.1D[3] -stim_base 7 -stim_label 7 dS -stim_file 8 motion_demean.1D[4] -stim_base 8 -stim_label 8 dL -stim_file 9 motion_demean.1D[5] -stim_base 9 -stim_label 9 dP -jobs 4 -fout -tout -x1D X.xmat.1D -xjpeg X.jpg -x1D_uncensored X.nocensor.xmat.1D -fitts fitts.01 -errts errts.01 -bucket stats.01
++ '-stim_times_AM1 1 ../stimuli/01SS.1D' has 1 auxiliary values per time point
++ '-stim_times_AM1 1': basis function model 'dmBLOCK' uses 1 parameters,
out of the 1 found in timing file '../stimuli/01SS.1D'
++ '-stim_times_AM1 2 ../stimuli/01LL.1D' has 1 auxiliary values per time point
++ '-stim_times_AM1 2': basis function model 'dmBLOCK' uses 1 parameters,
out of the 1 found in timing file '../stimuli/01LL.1D'
++ '-stim_times_AM2 3 ../stimuli/01stimlist2.1D' has 4 auxiliary values per time point
++ '-stim_times_AM2 3': basis function model 'dmBLOCK' uses 1 parameters,
out of the 4 found in timing file '../stimuli/01stimlist2.1D'
++ '-stim_times_AM2 3 ../stimuli/01stimlist2.1D' will have 4 regressors
++ 3dDeconvolve: AFNI version=AFNI_2011_12_21_1014 (Nov 23 2012) [64-bit]
++ Authored by: B. Douglas Ward, et al.
++ loading dataset ../pb03.01.r01.scale+tlrc.HEAD ../pb03.01.r02.scale+tlrc.HEAD ../pb03.01.r03.scale+tlrc.HEAD ../pb03.01.r04.scale+tlrc.HEAD
++ Auto-catenated input datasets treated as multiple imaging runs
++ Auto-catenated datasets start at: 0 180 360 540
++ STAT automask has 165684 voxels (out of 172800 = 95.9%)
++ Skipping check for initial transients
++ Input polort=3; Longest run=360.0 s; Recommended minimum polort=3 ++ OK ++
++ -stim_times using TR=2 s for stimulus timing conversion
++ -stim_times using TR=2 s for any -iresp output datasets
++ [you can alter the -iresp TR via the -TR_times option]
++ -stim_times_AM1 1 using GLOBAL times
++ -stim_times_AM1 2 using GLOBAL times
++ -stim_times_AM2 3 using GLOBAL times
++ '-stim_times_AM2 3' average amplitude#1=75.0025
++ '-stim_times_AM2 3' average amplitude#2=0.902089
++ '-stim_times_AM2 3' average amplitude#3=4.0155
++ Number of time points: 720 (before censor) ; 607 (after)
+ Number of parameters: 28 [22 baseline ; 6 signal]
++ total shared memory needed = 1,006,387,200 bytes (about 1.0 billion)
++ mmap() memory allocated: 1,006,387,200 bytes (about 1.0 billion)
++ Memory required for output bricks = 1,006,387,200 bytes (about 1.0 billion)
++ Wrote matrix image to file X.jpg
++ Wrote matrix values to file X.xmat.1D
++ ========= Things you can do with the matrix file =========
++ (a) Linear regression with ARMA(1,1) modeling of serial correlation:

3dREMLfit -matrix X.xmat.1D \
-input "../pb03.01.r01.scale+tlrc.HEAD ../pb03.01.r02.scale+tlrc.HEAD ../pb03.01.r03.scale+tlrc.HEAD ../pb03.01.r04.scale+tlrc.HEAD" \
-fout -tout -Rbuck stats.01_REML -Rvar stats.01_REMLvar \
-Rfitts fitts.01_REML -Rerrts errts.01_REML -verb

++ N.B.: 3dREMLfit command above written to file stats.REML_cmd
++ (b) Visualization/analysis of the matrix via ExamineXmat.R
++ (c) Synthesis of sub-model datasets using 3dSynthesize
++ ==========================================================
++ Wrote matrix values to file X.nocensor.xmat.1D
++ ----- Signal+Baseline matrix condition [X] (607x28): 4.71222 ++ VERY GOOD ++
*+ WARNING: !! in Signal+Baseline matrix:
* Largest singular value=2.1055
* 1 singular value is less than cutoff=2.1055e-07
* Implies strong collinearity in the matrix columns!
++ Signal+Baseline matrix singular values:
0 0.0948211 0.187422 0.233463 0.350705
0.450539 0.461093 0.638889 0.684352 0.793918
0.856242 0.892046 0.964599 0.970663 0.989858
0.996702 0.99857 1.01247 1.02003 1.037
1.05002 1.08233 1.11562 1.24859 1.37877
1.4562 1.67657 2.1055
++ ----- Signal-only matrix condition [X] (607x6): 1.42042 ++ VERY GOOD ++
*+ WARNING: !! in Signal-only matrix:
* Largest singular value=1.42799
* 1 singular value is less than cutoff=1.42799e-07
* Implies strong collinearity in the matrix columns!
++ Signal-only matrix singular values:
9.27832e-09 0.707764 0.857196 0.936871 1.3592
1.42799
++ ----- Baseline-only matrix condition [X] (607x22): 4.64398 ++ VERY GOOD ++
++ ----- stim_base-only matrix condition [X] (607x6): 2.66816 ++ VERY GOOD ++
++ ----- polort-only matrix condition [X] (607x16): 1.14773 ++ VERY GOOD ++
*+ WARNING: +++++ !! Matrix inverse average error = 0.00306919 ** BEWARE **
++ Matrix setup time = 1.54 s
** ERROR: !! 3dDeconvolve: Can't run past 3 matrix warnings without '-GOFORIT 3'
** ERROR: !! Currently at -GOFORIT 0
** ERROR: !! See file 3dDeconvolve.err for all WARNING and ERROR messages !!
** ERROR: !! Be sure you understand what you are doing before using -GOFORIT !!
** ERROR: !! If in doubt, consult with someone or with the AFNI message board !!
** FATAL ERROR: !! 3dDeconvolve (regretfully) shuts itself down !!
** Program compile date = Nov 23 2012