Divide by zero in entropy calculations about pypte HOT 7 CLOSED

One more thought on validation: Significance tests should be performed on the transfer entropies. Perhaps one could rather think in terms of bayesian inference rather than frequentist approaches. But that can be part of another project.

On PTE calculation: a proper choice of delay calculation is critical. I will refactor the code in a manner that the user will be aware of this issue.

from pypte.

Hi, thanks for the issue.

Can you provide some sample data so I can reproduce the warning?

from pypte.

I just ran mne_demo.py.

from pypte.

Oh okay. Will take a look on it soon. The code is overdue for refactoring anyway.

Pull requests are welcome :)

Divisions by zero will be ignored. For what reason would you like to tamper with your data to avoid the warning?

from pypte.

If it's indeed ignored, then there is no need.

By the way, I see you have some neural models implemented. Have you used them to validate the PTE here yet?

from pypte.

Yes. Otherwise it would be nonsense to apply it on real data sets.

I can open a gitter channel if you need any support.

from pypte.

I plan to use this for a couple of projects in the near future, so that might be helpful.

from pypte.