Incorrect decomposition by cvxEDA

[JiaerCGuo]

Hello all,

I'm new to Neurokit2 and cvxEDA so I would really appreciate some help here. I'm running eda_process on an EDA data downsampled by a factor of 20 from 500Hz to 25Hz. Here is the code I'm using to process:

signals, info = nk.eda_process(data_ds_event['EDA'], sampling_rate=25,
                              method_cleaning='neurokit',
                              method_phasic='cvxeda',
                              method_peaks='nabian2018',
                              report = 'text')

When I put down sampling_rate=25, the cvxEDA method's decomposition seemed incorrect

If I changed it to sampling_rate=10, the cvxEDA decomposition became better

When I changed the decomposition method to highpass and kept the sampling_rate=25, the results seemed fine, so I'm assuming the problem is not on the dataset itself.

Does anyone know what might be the problem here? I also tried to downsample the data to 10Hz, but similar thing happened that cvxEDA doesn't work well with the correct sampling rate, but better when the defined rate is lower.

Attached is the data I'm working with in case it's helpful to troubleshoot. Thanks in advance for any insights into this issue!

Cecilia Guo

Update: file with correct time stamps
data_ds_event.csv

[DominiqueMakowski]

the cvxEDA method's decomposition seemed incorrect

What do you mean by that, it looks pretty alright to me

[JiaerCGuo]

This might be clearer with the overlapped raw and tonic components. The first half of the tonic seems poorly decomposed with peaks and troughs that shouldn't be there.

[DominiqueMakowski]

the tonic peaks are fairly low frequency, it does not look that bad IMO (but then I'm mostly interested in the phasic signal, not too sure when it becomes a really good tonic)