TENT results not consistent with GAMMA results

irepe · June 14, 2018, 8:34am

Dear AFNI experts,

I run two 3dDeconvolve commands, one using a GAMMA function and one using a TENT function (see commands below).

In order to inspect whether the output using the TENT function for a specific condition was consistent with the results computed using a GAMMA function, I’ve uploaded both the -iresp output (underlay) and the stats output (overlay). I’m assuming a coherency between the findings in the two analyses. That is, I’m assuming that for the significant voxels using the GAMMA function, I should get a BOLD response time course that roughly resembles a logarithmic function or a u-inversed shape. This does not seem to be the case (see attached pic).

Could you help with understanding what might be causing the inconsistency?
The text stimfiles for the TENT function consisted of one row per run with the event onsets in seconds. Is this the correct way of setting up the stimfiles?

Thank for your assistance and help!

All the best,
Irene

3dDeconvolve -float -jobs 4 -input {data}/C_sub{sub}.results/all_runs.C_sub${sub}+tlrc -concat ‘1D: 0 185’ -nfirst 0 -num_stimts 20
-CENSORTR {data}/C_sub{sub}.results/censor_C_sub${sub}_combined_2.1D
-polort A
-stim_file 1 {data}/C_sub{sub}.results/motion_demean.1D’[0]’ -stim_base 1 -stim_label 1 ‘roll’
-stim_file 2 {data}/C_sub{sub}.results/motion_demean.1D’[1]’ -stim_base 2 -stim_label 2 ‘pitch’
-stim_file 3 {data}/C_sub{sub}.results/motion_demean.1D’[2]’ -stim_base 3 -stim_label 3 ‘yaw’
-stim_file 4 {data}/C_sub{sub}.results/motion_demean.1D’[3]’ -stim_base 4 -stim_label 4 ‘SI’
-stim_file 5 {data}/C_sub{sub}.results/motion_demean.1D’[4]’ -stim_base 5 -stim_label 5 ‘LR’
-stim_file 6 {data}/C_sub{sub}.results/motion_demean.1D’[5]’ -stim_base 6 -stim_label 6 ‘PA’
-stim_file 7 {data}/C_sub{sub}.results/motion_deriv.1D’[0]’ -stim_base 7 -stim_label 7 ‘roll_02’
-stim_file 8 {data}/C_sub{sub}.results/motion_deriv.1D’[1]’ -stim_base 8 -stim_label 8 ‘pitch_02’
-stim_file 9 {data}/C_sub{sub}.results/motion_deriv.1D’[2]’ -stim_base 9 -stim_label 9 ‘yaw_02’
-stim_file 10 {data}/C_sub{sub}.results/motion_deriv.1D’[3]’ -stim_base 10 -stim_label 10 ‘dS_02’
-stim_file 11 {data}/C_sub{sub}.results/motion_deriv.1D’[4]’ -stim_base 11 -stim_label 11 ‘dL_02’
-stim_file 12 {data}/C_sub{sub}.results/motion_deriv.1D’[5]’ -stim_base 12 -stim_label 12 ‘dP_02’
-local_times
-stim_times_AM1 13 {stim}/C_sub{sub}/stimfiles/game/1_Q_other.txt ‘dmBLOCK’ -stim_label 13 ‘qo’
-stim_times_AM1 14 {stim}/C_sub{sub}/stimfiles/game/2_Q_self.txt ‘dmBLOCK’ -stim_label 14 ‘qs’
-stim_times_AM1 15 {stim}/C_sub{sub}/stimfiles/game/3_C_other.txt ‘dmBLOCK’ -stim_label 15 ‘co’
-stim_times_AM1 16 {stim}/C_sub{sub}/stimfiles/game/4_C_self.txt ‘dmBLOCK’ -stim_label 16 ‘cs’
-stim_times_AM1 17 {stim}/C_sub{sub}/stimfiles/game/5_O_other_down.txt ‘dmBLOCK’ -stim_label 17 ‘ood’
-stim_times_AM1 18 {stim}/C_sub{sub}/stimfiles/game/6_O_other_up.txt ‘dmBLOCK’ -stim_label 18 ‘oou’
-stim_times_AM1 19 {stim}/C_sub{sub}/stimfiles/game/7_O_self_down.txt ‘dmBLOCK’ -stim_label 19 ‘osd’
-stim_times_AM1 20 {stim}/C_sub{sub}/stimfiles/game/8_O_self_up.txt ‘dmBLOCK’ -stim_label 20 ‘osu’
-gltsym ‘SYM: +qo -qs’
-glt_label 1 QS_vs_QO
-gltsym ‘SYM: +cs -co’
-glt_label 2 CS_vs_CO
-gltsym ‘SYM: +osd -ood’
-glt_label 3 OSD_vs_OOD
-gltsym ‘SYM: +osu -oou’
-glt_label 4 OSU_vs_OOU
-gltsym ‘SYM: +osd -osu’
-glt_label 5 OSD_vs_OSU
-gltsym ‘SYM: +osu -osd’
-glt_label 6 OSU_vs_OSD
-allzero_OK
-GOFORIT 6
-jobs 10
-fout -tout -x1D X.xmat.1D -xjpeg X.jpg
-x1D_uncensored X.nocensor.xmat.1D
-fitts fitts.subj \ -errts errts.{subj}
-bucket stats.C_sub${sub}_GAMMA \

3dDeconvolve -float -jobs 4 -input {data}/C_sub{sub}.results/all_runs.C_sub${sub}+tlrc -concat ‘1D: 0 185’ -nfirst 0 -num_stimts 20
-CENSORTR {data}/C_sub{sub}.results/censor_C_sub${sub}_combined_2.1D
-polort A
-stim_file 1 {data}/C_sub{sub}.results/motion_demean.1D’[0]’ -stim_base 1 -stim_label 1 ‘roll’
-stim_file 2 {data}/C_sub{sub}.results/motion_demean.1D’[1]’ -stim_base 2 -stim_label 2 ‘pitch’
-stim_file 3 {data}/C_sub{sub}.results/motion_demean.1D’[2]’ -stim_base 3 -stim_label 3 ‘yaw’
-stim_file 4 {data}/C_sub{sub}.results/motion_demean.1D’[3]’ -stim_base 4 -stim_label 4 ‘SI’
-stim_file 5 {data}/C_sub{sub}.results/motion_demean.1D’[4]’ -stim_base 5 -stim_label 5 ‘LR’
-stim_file 6 {data}/C_sub{sub}.results/motion_demean.1D’[5]’ -stim_base 6 -stim_label 6 ‘PA’
-stim_file 7 {data}/C_sub{sub}.results/motion_deriv.1D’[0]’ -stim_base 7 -stim_label 7 ‘roll_02’
-stim_file 8 {data}/C_sub{sub}.results/motion_deriv.1D’[1]’ -stim_base 8 -stim_label 8 ‘pitch_02’
-stim_file 9 {data}/C_sub{sub}.results/motion_deriv.1D’[2]’ -stim_base 9 -stim_label 9 ‘yaw_02’
-stim_file 10 {data}/C_sub{sub}.results/motion_deriv.1D’[3]’ -stim_base 10 -stim_label 10 ‘dS_02’
-stim_file 11 {data}/C_sub{sub}.results/motion_deriv.1D’[4]’ -stim_base 11 -stim_label 11 ‘dL_02’
-stim_file 12 {data}/C_sub{sub}.results/motion_deriv.1D’[5]’ -stim_base 12 -stim_label 12 ‘dP_02’
-local_times
-stim_times 13 {stim}/C_sub{sub}/stimfiles/game_tent/1_Q_other.txt ‘TENT(0,12,7)’ -stim_label 13 ‘qo’
-stim_times 14 {stim}/C_sub{sub}/stimfiles/game_tent/2_Q_self.txt ‘TENT(0,12,7)’ -stim_label 14 ‘qs’
-stim_times 15 {stim}/C_sub{sub}/stimfiles/game_tent/3_C_other.txt ‘TENT(0,12,7)’ -stim_label 15 ‘co’
-stim_times 16 {stim}/C_sub{sub}/stimfiles/game_tent/4_C_self.txt ‘TENT(0,12,7)’ -stim_label 16 ‘cs’
-stim_times 17 {stim}/C_sub{sub}/stimfiles/game_tent/5_O_other_down.txt ‘TENT(0,12,7)’ -stim_label 17 ‘ood’
-stim_times 18 {stim}/C_sub{sub}/stimfiles/game_tent/6_O_other_up.txt ‘TENT(0,12,7)’ -stim_label 18 ‘oou’
-stim_times 19 {stim}/C_sub{sub}/stimfiles/game_tent/7_O_self_down.txt ‘TENT(0,12,7)’ -stim_label 19 ‘osd’
-stim_times 20 {stim}/C_sub{sub}/stimfiles/game_tent/8_O_self_up.txt ‘TENT(0,12,7)’ -stim_label 20 ‘osu’
-iresp 13 iresp_qo
-iresp 14 iresp_qs
-iresp 15 iresp_co
-iresp 16 iresp_cs
-iresp 17 iresp_ood
-iresp 18 iresp_oou
-iresp 19 iresp_osd
-iresp 20 iresp_osu
-bucket stats.C_sub${sub}_TENT \

Gang · June 14, 2018, 8:54pm

Since TENT does not make any assumption about the shape of BOLD response, it may pick up something that is unrelated to BOLD signal. To diagnose the problem, add option -fitts to your TENT script (as you did for dmBLOCK), and show the fitted data versus the input file and see what is happening.

Also, I would change TENT to TENTzero to avoid picking up something at the onset times.

irepe · June 15, 2018, 10:24am

Dear Gang,

Thanks for your prompt reply and help.

I did what you suggested. I’m attaching the graphs with fitts (black) and all_runs (my input data, in green) overlayed. Do you have any insights?

Thank you!

All the best,
Irene

rickr · June 15, 2018, 5:55pm

Hi Irene,

It seems a little odd to compare dmBLOCK with TENT.
What is the range of duration values associated with
the events?

rick

irepe · July 19, 2018, 12:18pm

Dear Rick,

Thanks for your email.

The trials lasted either 8, 9 or 10 seconds. More specifically, each regressor lasted the following:
-stim_times_AM1 13 ${stim}/… ‘dmBLOCK’ -stim_label 13 ‘qo’ \ ----> duration 3 seconds
-stim_times_AM1 14 ${stim}/… ‘dmBLOCK’ -stim_label 14 ‘qs’ \ ----> 3 seconds
-stim_times_AM1 15 ${stim}/… ‘dmBLOCK’ -stim_label 15 ‘co’ \ ----> 2, 3 or 4 seconds
-stim_times_AM1 16 ${stim}/… ‘dmBLOCK’ -stim_label 16 ‘cs’ \ ----> 2, 3 or 4 seconds
-stim_times_AM1 17 ${stim}/… ‘dmBLOCK’ -stim_label 17 ‘ood’ \ ----> 3 seconds
-stim_times_AM1 18 ${stim}/… ‘dmBLOCK’ -stim_label 18 ‘oou’ \ ----> 3 seconds
-stim_times_AM1 19 ${stim}/… ‘dmBLOCK’ -stim_label 19 ‘osd’ \ ----> 3 seconds
-stim_times_AM1 20 ${stim}/… ‘dmBLOCK’ -stim_label 20 ‘osu’ \ ----> 3 seconds

Could you explain why one should not expect some consistency between dmBLOCK and TENT?

Thanks for your help!
Irene

rickr · July 20, 2018, 1:04pm

Hi Irene,

That is not too much variance in stimulus duration time,
so using TENTs would not be so bad, and that is only for
the ‘co’ and ‘cs’ classes, the others do not even vary.

The potential trouble with expecting TENT to be consistent
with dmBLOCK is that with dmBLOCK, one is usually meaning
that the stimulus duration will vary, and therefore the
shape of the response will vary. But when using TENT
to model the shape, it is not expected to vary.

In this case, with the only variance being between 2, 3
and 4 s, that is probably consistent enough to get
something reasonable with TENT, too.

Getting to your initial question then, the inconsistency
is probably due to having many events that overlap too
much. So the TENTs might be at least somewhat overmodeling
the data. Not only should events be more spread out, but
they should not tend to overlap in any predictable way.

How many time points are there? How many regressors in
the TENT case? How “random” is the timing and ISI?

rick

irepe · July 24, 2018, 8:14am

Dear Rick,

Thanks for your help and explanation, it really helps.

-How many time points are there?
Each trial lasted 8, 9 or 10 seconds and was divided in 3 within-trial events: “q” (3 sec), “c” (2, 3 or 4 sec) , “o” (3 sec).
For event “q” we had two possibilities/regressors (“qs” or “qo”), same for regressor “c” and for event “o” we had 4 possibilities/regressors (“ood”, “oou”, “osd”, “osu”). We had 32 trials per run (2 runs in total). TR=2.

-How many regressors in the TENT case?
I added as many regressors as I had for the dmBLOCK because it was my understanding that each within-trial event needed to be included (even though I’m mostly interested in what happened during event “o”). The text stimfiles for the TENT function consisted of one row per run with the event onsets in seconds. Here how it looks:

-stim_times 13 ${stim}/… ‘TENTzero(0,12,7)’ -stim_label 13 ‘qo’
-stim_times 14 ${stim}/… ‘TENTzero(0,12,7)’ -stim_label 14 ‘qs’
-stim_times 15 ${stim}/… ‘TENTzero(0,12,7)’ -stim_label 15 ‘co’
-stim_times 16 ${stim}/… ‘TENTzero(0,12,7)’ -stim_label 16 ‘cs’
-stim_times 17 ${stim}/… ‘TENTzero(0,12,7)’ -stim_label 17 ‘ood’
-stim_times 18 ${stim}/… ’TENTzero(0,12,7)’ -stim_label 18 ‘oou’
-stim_times 19 ${stim}/… ’TENTzero(0,12,7)’ -stim_label 19 ‘osd’
-stim_times 20 ${stim}/… ‘TENTzero(0,12,7)’ -stim_label 20 ‘osu’ \

Do you think that the duration was defined incorrectly? Should the duration relate to each regressor duration instead of the whole trial duration?

How “random” is the timing and ISI?
ISI was either 2, 3 or 4 seconds.

Thanks for your assistance!
Irene

rickr · July 24, 2018, 5:24pm

Hi Irene,

Had there been no jitter in ‘c’ duration, this would lead to
multi-collinearity problems with many sets of regressors.
While having the jitter may make this mathematically solvable,
the results should not be very robust. What were the matrix
warnings output from afni_proc.py? There were probably many
high correlations.

Basically, in order to use TENT effectively, the conditions
should not follow each other so consistently.

rick

irepe · July 25, 2018, 9:58am

Hi Rick,

Ok, I understand, thanks for your explanation.

The collinaerity is high between the regressors related to the “q” and “c” events.
For the regressors during the “o” event there’s a medium collinearity only with the motor regressor (see below).

Am I correct in assuming that the results for the “o” regressors are more solid and can be trusted?
Interestingly for those regressor I do find similarities between the results using GAMMA and TENT functions.(see attachments)

One last question regarding the graph of the iresp file. The TENT function I used included 12 seconds.
My regressor duration is 3 seconds. I should expect the first third of the graph (0-3 seconds) to reflect the time course related to the regressor. Correct? Or should I “shift” the expected response according to hemodynamic lag?

Thank you!
Irene

++ Smallest FDR q [0 Full_Fstat] = 6.1433e-14
++ Smallest FDR q [2 Q_other#0_Tstat] = 0.0301622
*+ WARNING: Smallest FDR q [4 Q_other#1_Tstat] = 0.538132 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [6 Q_other#2_Tstat] = 0.458426 ==> few true single voxel detections
++ Smallest FDR q [8 Q_other#3_Tstat] = 0.0521005
++ Smallest FDR q [10 Q_other#4_Tstat] = 0.0152592
*+ WARNING: Smallest FDR q [12 Q_other#5_Tstat] = 0.322586 ==> few true single voxel detections
++ Smallest FDR q [14 Q_other#6_Tstat] = 0.00761529
*+ WARNING: Smallest FDR q [15 Q_other_Fstat] = 0.334435 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [17 Q_self#0_Tstat] = 0.999789 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [19 Q_self#1_Tstat] = 0.231243 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [21 Q_self#2_Tstat] = 0.879093 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [23 Q_self#3_Tstat] = 0.103693 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [25 Q_self#4_Tstat] = 0.128768 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [27 Q_self#5_Tstat] = 0.250597 ==> few true single voxel detections
++ Smallest FDR q [29 Q_self#6_Tstat] = 0.0291363
++ Smallest FDR q [30 Q_self_Fstat] = 0.0609256
++ Smallest FDR q [32 C_other#0_Tstat] = 0.0361373
*+ WARNING: Smallest FDR q [34 C_other#1_Tstat] = 0.468286 ==> few true single voxel detections
++ Smallest FDR q [36 C_other#2_Tstat] = 0.050969
++ Smallest FDR q [38 C_other#3_Tstat] = 0.0113919
*+ WARNING: Smallest FDR q [40 C_other#4_Tstat] = 0.287053 ==> few true single voxel detections
++ Smallest FDR q [42 C_other#5_Tstat] = 0.00321686
*+ WARNING: Smallest FDR q [44 C_other#6_Tstat] = 0.457179 ==> few true single voxel detections
++ Smallest FDR q [45 C_other_Fstat] = 0.072442
*+ WARNING: Smallest FDR q [47 C_self#0_Tstat] = 0.999884 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [49 C_self#1_Tstat] = 0.917802 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [51 C_self#2_Tstat] = 0.103763 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [53 C_self#3_Tstat] = 0.134289 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [55 C_self#4_Tstat] = 0.266174 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [57 C_self#5_Tstat] = 0.941616 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [59 C_self#6_Tstat] = 0.885314 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [60 C_self_Fstat] = 0.999862 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [62 O_other_dwn#0_Tstat] = 0.999884 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [64 O_other_dwn#1_Tstat] = 0.497898 ==> few true single voxel detections
++ Smallest FDR q [66 O_other_dwn#2_Tstat] = 8.9075e-06
++ Smallest FDR q [68 O_other_dwn#3_Tstat] = 2.77526e-07
++ Smallest FDR q [70 O_other_dwn#4_Tstat] = 9.4016e-05
*+ WARNING: Smallest FDR q [72 O_other_dwn#5_Tstat] = 0.999636 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [74 O_other_dwn#6_Tstat] = 0.834283 ==> few true single voxel detections
++ Smallest FDR q [75 O_other_dwn_Fstat] = 2.81112e-06
++ Smallest FDR q [77 O_other_up#0_Tstat] = 0.0228593
*+ WARNING: Smallest FDR q [79 O_other_up#1_Tstat] = 0.367025 ==> few true single voxel detections
++ Smallest FDR q [81 O_other_up#2_Tstat] = 1.68745e-05
++ Smallest FDR q [83 O_other_up#3_Tstat] = 2.86516e-07
++ Smallest FDR q [85 O_other_up#4_Tstat] = 0.00029191
*+ WARNING: Smallest FDR q [87 O_other_up#5_Tstat] = 0.674601 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [89 O_other_up#6_Tstat] = 0.841764 ==> few true single voxel detections
++ Smallest FDR q [90 O_other_up_Fstat] = 9.01099e-06
*+ WARNING: Smallest FDR q [92 O_self_dwn#0_Tstat] = 0.579646 ==> few true single voxel detections
++ Smallest FDR q [94 O_self_dwn#1_Tstat] = 0.0994747
++ Smallest FDR q [96 O_self_dwn#2_Tstat] = 0.000646114
++ Smallest FDR q [98 O_self_dwn#3_Tstat] = 8.15666e-05
*+ WARNING: Smallest FDR q [100 O_self_dwn#4_Tstat] = 0.287892 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [102 O_self_dwn#5_Tstat] = 0.971845 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [104 O_self_dwn#6_Tstat] = 0.995507 ==> few true single voxel detections
++ Smallest FDR q [105 O_self_dwn_Fstat] = 2.82982e-05
*+ WARNING: Smallest FDR q [107 O_self_up#0_Tstat] = 0.690033 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [109 O_self_up#1_Tstat] = 0.121032 ==> few true single voxel detections
++ Smallest FDR q [111 O_self_up#2_Tstat] = 0.00201695
++ Smallest FDR q [113 O_self_up#3_Tstat] = 0.000798469
++ Smallest FDR q [115 O_self_up#4_Tstat] = 0.0559347
++ Smallest FDR q [117 O_self_up#5_Tstat] = 0.0632829
*+ WARNING: Smallest FDR q [119 O_self_up#6_Tstat] = 0.999896 ==> few true single voxel detections
++ Smallest FDR q [120 O_self_up_Fstat] = 5.45198e-06
*+ WARNING: Smallest FDR q [122 motor#0_Tstat] = 0.367716 ==> few true single voxel detections
++ Smallest FDR q [124 motor#1_Tstat] = 0.0298197
++ Smallest FDR q [126 motor#2_Tstat] = 0.00228571
++ Smallest FDR q [128 motor#3_Tstat] = 0.00241847
++ Smallest FDR q [130 motor#4_Tstat] = 0.0842148
*+ WARNING: Smallest FDR q [132 motor#5_Tstat] = 0.812234 ==> few true single voxel detections
*+ WARNING: Smallest FDR q [134 motor#6_Tstat] = 0.579016 ==> few true single voxel detections
++ Smallest FDR q [135 motor_Fstat] = 0.0531847

Warnings regarding Correlation Matrix: X.xmat.1D
severity correlation cosine regressor pair

[size=x-small]high: 1.000 1.000 (10 vs. 23) Q_other#2 vs.C_other#1
high: 1.000 1.000 (11 vs. 24) Q_other#3 vs.C_other#2
high: 1.000 1.000 (12 vs. 25) Q_other#4 vs.C_other#3
high: 1.000 1.000 (13 vs. 26) Q_other#5 vs.C_other#4
high: 1.000 1.000 (20 vs. 33) Q_self#5 vs.C_self#4
high: 1.000 1.000 (19 vs. 32) Q_self#4 vs.C_self#3
high: 1.000 1.000 (18 vs. 31) Q_self#3 vs.C_self#2
high: 1.000 1.000 (17 vs. 30) Q_self#2 vs.C_self#1
high: 0.814 0.825 (22 vs. 65) C_other#0 vs.motor#1
high: 0.781 0.797 (27 vs. 70) C_other#5 vs.motor#6
high: 0.744 0.764 (14 vs. 69) Q_other#6 vs.motor#5
high: 0.732 0.752 ( 9 vs. 64) Q_other#1 vs.motor#0
high: 0.731 0.752 (14 vs. 27) Q_other#6 vs.C_other#5
high: 0.730 0.765 (26 vs. 69) C_other#4 vs.motor#5
high: 0.730 0.765 (25 vs. 68) C_other#3 vs.motor#4
high: 0.730 0.765 (24 vs. 67) C_other#2 vs.motor#3
high: 0.730 0.765 (23 vs. 66) C_other#1 vs.motor#2
high: 0.720 0.756 (13 vs. 69) Q_other#5 vs.motor#5
high: 0.720 0.756 (12 vs. 68) Q_other#4 vs.motor#4
high: 0.720 0.756 (11 vs. 67) Q_other#3 vs.motor#3
high: 0.720 0.756 (10 vs. 66) Q_other#2 vs.motor#2
high: 0.720 0.756 ( 9 vs. 65) Q_other#1 vs.motor#1
medium: 0.693 0.717 (21 vs. 34) Q_self#6 vs.C_self#5
medium: 0.624 0.673 (13 vs. 68) Q_other#5 vs.motor#4
medium: 0.624 0.673 (12 vs. 67) Q_other#4 vs.motor#3
medium: 0.624 0.673 (11 vs. 66) Q_other#3 vs.motor#2
medium: 0.624 0.673 (10 vs. 65) Q_other#2 vs.motor#1
medium: 0.622 0.652 (16 vs. 29) Q_self#1 vs.C_self#0
medium: 0.618 0.648 ( 9 vs. 22) Q_other#1 vs.C_other#0
medium: 0.617 0.649 ( 9 vs. 35) Q_other#1 vs.C_self#6
medium: 0.614 0.664 (27 vs. 69) C_other#5 vs.motor#5
medium: 0.614 0.664 (26 vs. 68) C_other#4 vs.motor#4
medium: 0.614 0.664 (25 vs. 67) C_other#3 vs.motor#3
medium: 0.614 0.664 (24 vs. 66) C_other#2 vs.motor#2
medium: 0.614 0.664 (23 vs. 65) C_other#1 vs.motor#1
medium: 0.603 0.642 (39 vs. 69) O_other_dwn#3 vs.motor#5
medium: 0.603 0.642 (38 vs. 68) O_other_dwn#2 vs.motor#4
medium: 0.603 0.642 (37 vs. 67) O_other_dwn#1 vs.motor#3
medium: 0.594 0.634 (30 vs. 41) C_self#1 vs.O_other_dwn#5
medium: 0.592 0.632 (17 vs. 41) Q_self#2 vs.O_other_dwn#5
medium: 0.592 0.632 (16 vs. 40) Q_self#1 vs.O_other_dwn#4
medium: 0.581 0.613 (14 vs. 39) Q_other#6 vs.O_other_dwn#3
medium: 0.579 0.611 (28 vs. 40) C_other#6 vs.O_other_dwn#4
medium: 0.564 0.602 (16 vs. 28) Q_self#1 vs.C_other#6
medium: 0.557 0.595 (35 vs. 65) C_self#6 vs.motor#1
medium: 0.544 0.590 (12 vs. 37) Q_other#4 vs.O_other_dwn#1
medium: 0.544 0.590 (13 vs. 38) Q_other#5 vs.O_other_dwn#2
medium: 0.539 0.585 (25 vs. 37) C_other#3 vs.O_other_dwn#1
medium: 0.539 0.585 (26 vs. 38) C_other#4 vs.O_other_dwn#2
medium: 0.539 0.585 (27 vs. 39) C_other#5 vs.O_other_dwn#3
medium: 0.525 0.565 (15 vs. 69) Q_self#0 vs.motor#5
medium: 0.524 0.554 ( 8 vs. 60) Q_other#0 vs.O_self_up#3
medium: 0.524 0.554 (22 vs. 61) C_other#0 vs.O_self_up#4
medium: 0.516 0.547 (21 vs. 53) Q_self#6 vs.O_self_dwn#3
medium: 0.515 0.555 ( 8 vs. 34) Q_other#0 vs.C_self#5
medium: 0.513 0.554 (62 vs. 66) O_self_up#5 vs.motor#2
medium: 0.513 0.554 (61 vs. 65) O_self_up#4 vs.motor#1
medium: 0.512 0.553 (15 vs. 27) Q_self#0 vs.C_other#5
medium: 0.496 0.518 (18 vs. 50) Q_self#3 vs.O_self_dwn#0
medium: 0.494 0.515 (31 vs. 50) C_self#2 vs.O_self_dwn#0
medium: 0.493 0.536 (20 vs. 52) Q_self#5 vs.O_self_dwn#2
medium: 0.493 0.536 (19 vs. 51) Q_self#4 vs.O_self_dwn#1
medium: 0.490 0.523 (35 vs. 54) C_self#6 vs.O_self_dwn#4
medium: 0.488 0.531 (34 vs. 53) C_self#5 vs.O_self_dwn#3
medium: 0.488 0.531 (33 vs. 52) C_self#4 vs.O_self_dwn#2
medium: 0.488 0.531 (32 vs. 51) C_self#3 vs.O_self_dwn#1
medium: 0.486 0.530 (19 vs. 58) Q_self#4 vs.O_self_up#1
medium: 0.486 0.530 (20 vs. 59) Q_self#5 vs.O_self_up#2
medium: 0.480 0.525 (32 vs. 58) C_self#3 vs.O_self_up#1
medium: 0.480 0.525 (33 vs. 59) C_self#4 vs.O_self_up#2
medium: 0.480 0.525 (34 vs. 60) C_self#5 vs.O_self_up#3
medium: 0.479 0.512 ( 8 vs. 21) Q_other#0 vs.Q_self#6
medium: 0.476 0.507 (11 vs. 36) Q_other#3 vs.O_other_dwn#0
medium: 0.475 0.506 (24 vs. 36) C_other#2 vs.O_other_dwn#0
medium: 0.470 0.501 (36 vs. 66) O_other_dwn#0 vs.motor#2
medium: 0.461 0.489 (18 vs. 57) Q_self#3 vs.O_self_up#0
medium: 0.457 0.486 (31 vs. 57) C_self#2 vs.O_self_up#0
medium: 0.457 0.495 (44 vs. 67) O_other_up#1 vs.motor#3
medium: 0.457 0.495 (45 vs. 68) O_other_up#2 vs.motor#4
medium: 0.457 0.495 (46 vs. 69) O_other_up#3 vs.motor#5
medium: 0.453 0.488 (22 vs. 35) C_other#0 vs.C_self#6
medium: 0.451 0.487 (35 vs. 61) C_self#6 vs.O_self_up#4
medium: 0.447 0.495 (23 vs. 62) C_other#1 vs.O_self_up#5
medium: 0.444 0.492 (10 vs. 62) Q_other#2 vs.O_self_up#5
medium: 0.444 0.492 ( 9 vs. 61) Q_other#1 vs.O_self_up#4
medium: 0.439 0.470 (47 vs. 70) O_other_up#4 vs.motor#6
medium: 0.436 0.473 (28 vs. 29) C_other#6 vs.C_self#0
medium: 0.427 0.465 (21 vs. 60) Q_self#6 vs.O_self_up#3
medium: 0.425 0.500 (16 vs. 27) Q_self#1 vs.C_other#5
medium: 0.419 0.461 (15 vs. 39) Q_self#0 vs.O_other_dwn#3
medium: 0.417 0.468 (10 vs. 55) Q_other#2 vs.O_self_dwn#5
medium: 0.417 0.468 ( 9 vs. 54) Q_other#1 vs.O_self_dwn#4
medium: 0.416 0.466 (23 vs. 55) C_other#1 vs.O_self_dwn#5
medium: 0.411 0.471 (27 vs. 40) C_other#5 vs.O_other_dwn#4
medium: 0.411 0.471 (26 vs. 39) C_other#4 vs.O_other_dwn#3
medium: 0.411 0.471 (25 vs. 38) C_other#3 vs.O_other_dwn#2
medium: 0.411 0.471 (24 vs. 37) C_other#2 vs.O_other_dwn#1
medium: 0.410 0.449 (14 vs. 15) Q_other#6 vs.Q_self#0
medium: 0.404 0.465 (13 vs. 39) Q_other#5 vs.O_other_dwn#3
medium: 0.404 0.465 (12 vs. 38) Q_other#4 vs.O_other_dwn#2
medium: 0.404 0.465 (11 vs. 37) Q_other#3 vs.O_other_dwn#1
medium: 0.402 0.440 (54 vs. 64) O_self_dwn#4 vs.motor#0
medium: 0.400 0.433 (36 vs. 37) O_other_dwn#0 vs.O_other_dwn#1[/size]

rickr · July 25, 2018, 2:58pm

Hi Irene,

Yup, that is an impressive list of high correlations.
Note that the values of 1.0 should actually result in
regression failure, even a single one. That is
multi-collinearity, where there would be an infinite
number of equivalent solutions to the regression model.
Since 3dDeconvolve uses SVD, this is allowed, but still
bad.

The “o” correlations are lower, but still pretty high.
They are probably lower since they follow the variable
duration events. The actual numbers alone do not imply
whether it is problematic. It is the consistent overlap
with the adjacent conditions that is cause for concern.
The results will still be noisy, and may or may not look
like they should.

So it is difficult to say how trustworthy they are
(which I guess actually means they are not terribly
trustworthy :).

Also, since these conditions last about 3 seconds, one
would normally expect a 15+ second response curve, not
12, and you are already omitting the endpoints. That
may lead to a response shift of 1 or 2 TRs.

Since your TR seems to be 2, consider adding 1 or 2 of
them to the response curves.

In any case, TENT does not seem very appropriate here.

rick

irepe · July 30, 2018, 7:08am

Hi Rick,

THanks for your help and explanation!

Irene