dsfsi/covid19za

[Feature] Add a new Rt notebook with MCMC

vukosim opened this issue · 5 comments

@heerden could you assist and just have a 2nd rt file just denoted that it is using MCMC

@vukosim very excited that they released an updated MCMC model. I will review it this weekend.

To match the US Covid Tracking Project daily data set to our SA data sets, can you confirm if these fields represent the same meaning?
US positive -> total (covid19za_provincial_cumulative_timeline_confirmed.csv)
US total -> cumulative_tests (covid19za_timeline_testing.csv)

We will then be limited to only country-level Rt results with this multivariable model. What is the state of covid19za_provincial_cumulative_timeline_testing.csv? It does not seem to be updated regularly.

@vukosim The draft is available on my fork
https://github.com/heerden/covid-model/blob/master-za/rtlive-model-za.ipynb. Will format for export and add all the model files to our repo later in the week.

With the new model Rt seems very constant at 1.2 for the last month.

Thanks @heerden

Can you check the current challenge with the automation and also add the mcmc automation. LGTM

@vukosim Sorry, I should have followed up earlier on the last 3 automated action failures, but had another deadline.

Thanks for confirming the MCMC model.

The remote server is extremely finicky with pymc3 in a notebook, thus I will use a pure python script.

Will submit daily manual results for now and commit the automation on the weekend.