dsfsi/covid19za

[Feature] Resume Bayesian Rt calculations

vukosim opened this issue · 5 comments

vukosim commented

Could we just resume those Rt calculations for the Bayesian model @heerden

vhschalk commented

There is an open issue on the notebook automation action recipe, thus the action is failing. Will check up on this daily and manually commit for now.

vukosim commented

Ahh thanks @heerden

Hope you are well.

  • 👍1
vukosim commented

@heerden Quick question, we are no longer getting calculations for GP province in the main file. Is there are a reason for this?

vhschalk commented

@vukosim In the case of GP data, in the highest_density_interval functions, this sequence becomes zero: best = (highs - lows).argmin(). There is a try-except block to prevent errors like this and will ignore provinces with this error. This means that the confidence interval could not be calculated for GP. I explicitly removed the confidence interval from the SA data, that Rt for could at least be logged on the country level.

See my comments at line 154 here. Note that all the calculation has been moved to pure Python script, since the Notebook recipe (mentioned on 16 Sep), still does not work.

I cannot find a way to debug the model further, to always return a confidence interval.

We can however make to call to still just show the GP results again, without the confidence interval?

vukosim commented

Thanks. I will also investigate. And get back. THanks @heerden