opt is 1 1 1 1，but only modelA and ModelB were finished

[Jiaying214]

Hi I'm new in using GLMsingle.

I have encountered a couple of issues and would appreciate your assistance:

1. When using the toolbox, I followed the same settings as in example1. However, in the output directory, only model typeA and model typeB are computed and generated when the analysis is completed. I am unsure why this is happening. Below are the settings I used:

opt['wantlibrary'] = 1
opt['wantglmdenoise'] = 1
opt['wantfracridge'] = 1
opt['wantfileoutputs'] = [1,1,1,1]
opt['wantmemoryoutputs'] = [1,1,1,1]

The output is :

*** DIAGNOSTICS ***:
There are 1 runs.
The number of conditions in this experiment is 2.
The stimulus duration corresponding to each trial is 1.00 seconds.
The TR (time between successive data points) is 2.00 seconds.
The number of trials in each run is: [60].
The number of trials for each condition is: [30, 30].
For each condition, the number of runs in which it appears: [1, 1].
For each run, how much ending buffer do we have in seconds? [18].
*** Saving design-related results to /media/task/mvpa/singleglm/test/result/DESIGNINFO.npy. ***
*** FITTING DIAGNOSTIC RUN-WISE FIR MODEL ***
/home/qinlab/anaconda3/lib/python3.12/site-packages/glmsingle/glmsingle.py:655: UserWarning: None of your conditions occur in more than one run. Are you sure this is what you intend?
warnings.warn(msg)
/home/qinlab/anaconda3/lib/python3.12/site-packages/glmsingle/glmsingle.py:665: UserWarning: Since there are no repeats, standard cross-validation usage of cannot be performed. Setting to 0.
warnings.warn(msg)
/home/qinlab/anaconda3/lib/python3.12/site-packages/glmsingle/glmsingle.py:672: UserWarning: Since there are no repeats, standard cross-validation usage of cannot be performed. Setting to 0.
warnings.warn(msg)
*** Saving FIR results to /media/task/mvpa/singleglm/test/result/RUNWISEFIR.npy. ***

*** FITTING TYPE-A MODEL (ONOFF) ***

fitting model...
done.

preparing output...
done.

computing model fits...
done.

computing R^2...
done.

computing SNR...
done.

*** Saving results to /media/task/mvpa/singleglm/test/result/TYPEA_ONOFF.npy. ***

/home/qinlab/anaconda3/lib/python3.12/site-packages/sklearn/mixture/_base.py:270: ConvergenceWarning: Best performing initialization did not converge. Try different init parameters, or increase max_iter, tol, or check for degenerate data.
warnings.warn(
*** Setting brain R2 threshold to 0.161440295976968 ***

*** FITTING TYPE-B MODEL (FITHRF) ***

chunks: 100%|██████████| 7/7 [01:37<00:00, 13.88s/it]

*** Saving results to /media/task/mvpa/singleglm/test/result/TYPEB_FITHRF.npy. ***

*** All model types done ***

*** return model types in results ***

2. Could this toolbox be used for REST fMRI analysis? Additionally, is there an output file similar to the errts file generated by AFNI (using the command 3dDeconvolve)?

Thanks!!

[kendrickkay]

Re: 1 -> if you provide only one run, GLMsingle cannot do the GLMdenoise and ridge regression procedures; hence you receive only typeA and typeB. Re: 2 -> GLMsingle is not really intended for resting-state data since there is no explicit repeatable experimental manipulation. Regarding the errts (residuals), this would be a nice feature but it is not currently implemented in GLMsingle

…

Hi I'm new in using GLMsingle. I have encountered a couple of issues and would appreciate your assistance: When using the toolbox, I followed the same settings as in example1. However, in the output directory, only model typeA and model typeB are computed and generated when the analysis is completed. I am unsure why this is happening. Below are the settings I used: opt['wantlibrary'] = 1 opt['wantglmdenoise'] = 1 opt['wantfracridge'] = 1 opt['wantfileoutputs'] = [1,1,1,1] opt['wantmemoryoutputs'] = [1,1,1,1] The output is : *** DIAGNOSTICS ***: There are 1 runs. The number of conditions in this experiment is 2. The stimulus duration corresponding to each trial is 1.00 seconds. The TR (time between successive data points) is 2.00 seconds. The number of trials in each run is: [60]. The number of trials for each condition is: [30, 30]. For each condition, the number of runs in which it appears: [1, 1]. For each run, how much ending buffer do we have in seconds? [18]. *** Saving design-related results to /media/task/mvpa/singleglm/test/result/DESIGNINFO.npy. *** *** FITTING DIAGNOSTIC RUN-WISE FIR MODEL *** /home/qinlab/anaconda3/lib/python3.12/site-packages/glmsingle/glmsingle.py:655: UserWarning: None of your conditions occur in more than one run. Are you sure this is what you intend? warnings.warn(msg) /home/qinlab/anaconda3/lib/python3.12/site-packages/glmsingle/glmsingle.py:665: UserWarning: Since there are no repeats, standard cross-validation usage of cannot be performed. Setting to 0. warnings.warn(msg) /home/qinlab/anaconda3/lib/python3.12/site-packages/glmsingle/glmsingle.py:672: UserWarning: Since there are no repeats, standard cross-validation usage of cannot be performed. Setting to 0. warnings.warn(msg) *** Saving FIR results to /media/task/mvpa/singleglm/test/result/RUNWISEFIR.npy. *** *** FITTING TYPE-A MODEL (ONOFF) *** fitting model... done. preparing output... done. computing model fits... done. computing R^2... done. computing SNR... done. *** Saving results to /media/task/mvpa/singleglm/test/result/TYPEA_ONOFF.npy. *** /home/qinlab/anaconda3/lib/python3.12/site-packages/sklearn/mixture/_base.py:270: ConvergenceWarning: Best performing initialization did not converge. Try different init parameters, or increase max_iter, tol, or check for degenerate data. warnings.warn( *** Setting brain R2 threshold to 0.161440295976968 *** *** FITTING TYPE-B MODEL (FITHRF) *** chunks: 100%|██████████| 7/7 [01:37<00:00, 13.88s/it] *** Saving results to /media/task/mvpa/singleglm/test/result/TYPEB_FITHRF.npy. *** *** All model types done *** *** return model types in results *** Could this toolbox be used for REST fMRI analysis? Additionally, is there an output file similar to the errts file generated by AFNI (using the command 3dDeconvolve)? Thanks!! — Reply to this email directly, view it on GitHub <#181>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAU2DEULASEIKH4QB4JHKBL3DAIFDAVCNFSM6AAAAAB7CJF4LGVHI2DSMVQWIX3LMV43ERDJONRXK43TNFXW4OZYGQ2DCMRVHE>. You are receiving this because you are subscribed to this thread.

[Jiaying214]

Thanks！！

[SteveSizzou]

@kendrickkay thanks for the info. I have a similar set up to @Jiaying214 in that my condition repeats all take place within a single run and so I was also confused over why model D wasn't returned despite also using "opt['wantlibrary'] = 1, opt['wantglmdenoise'] = 1, opt['wantfracridge'] = 1".

Do you know, is it something inherent in GLMsingle that CV can only be appled across multiple runs, or could I artifically break my single run up into separate lists (given that I have 6 trials per each condition I would then break the single run into 6 separate "psuedo runs") in order to then perform CV?

Thanks in advance for your thoughts!

[kendrickkay]

Yes - the code is setup so that CV has to occur across distinct runs. (If this is not the case, the type C/D will not be run.) You could certainly try the approach of breaking a single run into several (pseudo-) runs. How well that approach will work depends on all sorts of factors like the number of time samples, whether the "run breaks" are interrupting real BOLD signals, etc.

…

-- Kendrick Kay ***@***.***) Associate Professor Center for Magnetic Resonance Research University of Minnesota Lab web site: http://cvnlab.net Online booking site: https://kendrickkay.youcanbook.me

On Aug 14, 2025, at 6:04 AM, SteveSizzou ***@***.***> wrote: @kendrickkay <https://github.com/kendrickkay> thanks for the info. I have a similar set up to @Jiaying214 <https://github.com/Jiaying214> in that my condition repeats all take place within a single run and so I was also confused over why model D wasn't returned despite also using "opt['wantlibrary'] = 1, opt['wantglmdenoise'] = 1, opt['wantfracridge'] = 1". Do you know, is it something inherent in GLMsingle that CV can only be appled across multiple runs, or could I artifically break my single run up into separate lists (given that I have 6 trials per each condition I would then break the single run into 6 separate "psuedo runs") in order to then perform CV? Thanks in advance for your thoughts! — Reply to this email directly, view it on GitHub <#181 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAU2DEXAUXOEUCTGBIAEFY33NRUKFAVCNFSM6AAAAAB7CJF4LGVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTIMJQGI3DCNQ>. You are receiving this because you were mentioned.