[Eeglablist] Increase running speed

[Makoto Miyakoshi]

Hello Marjan,

> And the problem I have now is that it takes a long time to run for a
subject, it took a day and it's still running, and I want to know if you
have any suggestions or solutions to help me increase my running speed?

If I remember correctly, the process of model order estimation for
Vieira-Morf has a weird bug and it does not finish the process. Once it
happens in one of the datasets, your whole pipeline is done!
To avoid this, I used to use different algorithms, such as ARfit.

A suggestion for speed up your process is to run statistics separately.
Your code has 200 permutations, which means you run the whole process (!)
200 times. I always used standard stats on my own, such as ttest2().

I also recommend resample to 100 Hz. Model order can be searched only up to
20 (practically, I would not use any results that show model order >= 20
from experience). I forgot what 'window length' means, but if it is in s, I
guess you run this on continuous data?

I would use 'freqs' 2:0.2:45 and use pseudo-log scale for the freq axis to
give it much better freq resolution below the alpha range.

Makoto


On Wed, Jan 8, 2025 at 5:17 AM m za <ma.zamani.20 at gmail.com> wrote:

> Hello,
>
> I want to  obtain  connectivity using MATLAB eeglab codes and the problem
> I had was that I was getting model order error, and with the help of the
> guidance I found in this link (
> https://sccn.ucsd.edu/pipermail/eeglablist/2018/013225.html), my problem
> was solved. And the problem I have now is that it takes a long time to run
> for a subject, it took a day and it's still running, and I want to know if
> you have any suggestions or solutions to help me increase my running speed?
> And whether the code I'm using is correct or needs changes?
>
> Thank you in advance for any help you can provide.
> Marjan
>
>
> %myfilename1=['E:\FinalCode\Thises\eeglaboutput\dir\dir_sub',num2str(q),'.set'];
> %EEG=load(myfilename1,'-mat');
> [ALLEEG EEG] = eeg_store(ALLEEG, EEG, CURRENTSET);
> EEG=pop_chanedit(EEG, 'setref',{'1:19','A1,A2'},'eval','chans =
> pop_chancenter( chans, [],[]);');
> EEG = pop_resample(EEG,250);
> %EEG = pop_epoch( EEG, { 'square' }, [0 5], 'epochinfo', 'yes')
> % prepare data
> EEG = pop_pre_prepData(EEG, 'nogui', 'SignalType',{'Channels'},
> 'NormalizeData', {'Method' {'time' 'ensemble'} }, 'verb', 1);
> %Remove Baseline
> EEG = pop_rmbase(EEG,1:19);
> % Model Order
> EEG.CAT.configs.est_selModelOrder = arg_tovals( arg_report( 'rich',
> @est_selModelOrder));
> EEG.CAT.configs.est_selModelOrder.modelingApproach.winlen =5;
> EEG = pop_est_selModelOrder(EEG, 'nogui', 'verb', 1,'prctWinToSample',
> 100);
> %EEG = pop_est_selModelOrder(EEG,
> 'nogui','ModelingApproach','SegmentationVAR','Algorithm',
> 'Vieira-Morf','WindowLength', 0.5, 'WindowStepSize',
> 0.032,'TaperFunction','rectwin','NormalizeData','True','DetrendingMethod','constant','ModelOrderRange',[1
> 30],'InformationCriteria',{'sbc' 'aic' 'fpe' 'hq'}, 'verb', 1);
> % fit AR model
> EEG = pop_est_fitMVAR(EEG, 'nogui', 'Algorithm', 'Vieira-Morf',
> 'ModelOrder', 30, 'WindowLength', 5, 'WindowStepSize', 0.032, 'verb', 1);
> %Validate Model
> %EEG = pop_est_validateMVAR(EEG,'nogui','whitenessCriteria',{'Ljung-Box'
> 'ACF' 'Box-Pierce'
> 'Li-McLeod'},'CheckResidualsWhiteness','True','SignificanceLevel',0.05,'CheckConsistency','True','CheckStability','True',
> 'verb', 1);
> EEG.CAT.configs.est_selModelOrder.CheckResidualsWhiteness='True';
> EEG = pop_est_validateMVAR(EEG,'nogui', 'verb', 1);
> % Compute connectivity
> EEG = pop_est_mvarConnectivity(EEG, 'nogui', 'ConnectivityMeasures', {'dDTF'
> 'GPDC' 'GGC'}, 'freqs', 0.5:45, 'verb', 1);
> % Compute surrogate
> %EEG.CAT.configs.est_selModelOrder.Mode={'PhaseRand'};
> EEG.CAT.configs.est_selModelOrder.NumPermutaions= 200;
> %EEG =
> pop_stat_surrogateGen(EEG,'nogui','ModelingApproach','SegmentationVAR','NormalizeData','True','Mode',{'PhaseRand'},'NumPermutaions'
> , 200 , 'verb', 1);
> %EEG = pop_stat_surrogateGen(EEG,'nogui','verb', 1);
> EEG = pop_stat_surrogateGen(EEG,'nogui','Mode',{'PhaseRand'},'verb', 1);
> EEG.CAT.configs.est_selModelOrder.StatisticalTest='Hnull';
> EEG.CAT.configs.est_selModelOrder.MultipleComparisonCorrection='fdr';
> %EEG =
> pop_stat_surrogateStats(EEG,'nogui','StatisticalTest','Hnull','MultipleComparisonCorrection','fdr',
> 'verb', 1);
> EEG = pop_stat_surrogateStats(EEG,'nogui','verb', 1);
> %pop_vis_TimeFreqGrid(EEG);
> %vis_TimeFreqGrid('EEG', EEG(1), 'Conn', EEG(1).CAT.Conn, 'MatrixLayout',
> ...
> %{'Partial', 'UpperTriangle', 'dDTF' 'LowerTriangle', 'dDTF' 'Diagonal'
> 'None' 'AllColorLimits' 99.9 }, ...
> %'baseline', [-1 -0.25], 'Thresholding', {'Simple', 'PercentileThreshold',
> [97.5 3]});
>