[Bug] Error when the number of Fatal cases are not changing with SIR-D model about covid19-sir HOT 10 OPEN

Thank you for your bug report!
However, I failed in reporducing the error with Google Colab and the same codes.
https://gist.github.com/lisphilar/e25b5ac4d1ac9dcea0f383ead0a4957e

Does the following scpript return empty dataframe?
(If yes, error is in downloading. If no, error is in subsetting.)

import covsirphy as cs
cs.__version__

eng = cs.DataEngineer()
eng.download(country='USA', province='Illinois', databases=["japan", "covid19dh"]);
eng.clean()
eng.transform()

df = eng.all()
df.loc[df['City'] == 'Cook']

from covid19-sir.

Thanks you for replying promptly!
Yes it does return a dataframe. I think I found the issue with my code. I didn't set the layer for the dataengineer

import covsirphy as cs
cs.version

eng = cs.DataEngineer(['ISO3', 'Province', 'City'], complement=True)
eng.download(country='USA', province='Illinois', databases=["japan", "covid19dh", "owid"])
eng.clean()
eng.transform()
eng.layer(geo=("USA", "Illinois",))

actual_df, status, _ = eng.subset(geo=("USA", 'Illinois', 'Cook'), variables="SIRF", complement=True)
print(status)
actual_df.tail()

This gives a subset on city level. However the sample code doesn't add any layers. Therefore eng.subset is acting funny if you provide geo=('Japan', 'Tokyo', ) or any other combination of city and province.

from covid19-sir.

I may not have understood the point correctly, but cs.DataEngineer() is the same as cs.DataEngineer(layers=None) and cs.DataEngineer(layers=[“ISO3”, “Province”, “City”]). Is this making you confused?

from covid19-sir.

I think I was able to solve the confusion.

I am trying to create a phase dependent model (SIRD) for a city in USA.

It seems that value of kappa is 0 for some phases.
I have tried different algo for segments and other arguments too. But still it is 0.
Can you suggest a work around.

Colab link below:
https://colab.research.google.com/drive/1dFKxUWd6L7ek_x1cognX49sChrE5142e#scrollTo=MD7lhKDtG6ZK

from covid19-sir.

Hi @karanjogi
I'm sorry for my very late reply.

I performed additional analysis and found that the number of fatal cases were not changed in the phases with kappa = 0.
https://gist.github.com/lisphilar/d717d73bcb1f5d3e80e79db9bbee4796

If the number of fatal cases is correct, kappa = 0 should be accepted. i.e. We need to fix the ZeroDivisionError by returning NA when kappa = 0.

from covid19-sir.

Hi @lisphilar ,

Thank you for clearing what the problem is!

Can you please suggest a workaround this error? Or where the code change needs to be done?
I can give it a try and submit a PR if solved.

Thanks!

from covid19-sir.

Thank you for your contribution!

Regarding SIR-D model, the error is raised here. Could you add if statements in "try" block?

The other models (sir.py, sirf.py, sewirf.py) in the directory also have the same issues.

Additionally, please add tests (with pytest) to https://github.com/lisphilar/covid19-sir/blob/5eb55efc36db79f302c951878d8bd45bf7e6e6dc/tests/test_dynamics/test_model.py so that we can avoid the similar issues.

I look forward to our discussion.

from covid19-sir.

And could you add a title to this issue?

from covid19-sir.

Sure!
I have made the changes in a new branch named develop and will be creating a PR for it.
Could you please review it and let me know if there are any changes needed?

from covid19-sir.

Thank you got your pull request!
As I commented on that, rho_i = 0 and sigma = 0 could be acceptable as well as kappa = 0.

from covid19-sir.