Closed djay closed 2 years ago
I added in this plot but didn't show things as clearly as I thought.
in theory the CFR should drop as vaccines kick in since less people die for the same number of cases. but it seems to effected too much by the detection rate cases maybe?
it also looks very different from OWID one. because I think they calculate it based on total deaths / total cases on each date. Whereas I'm calculating 14 avg deaths shifted 11 days) / 14 day avg cases to try and get a rough daily figure. So I probably need to call it somthing else.
I think using the OWID way is perhaps better even if it going to show less and less variation over time. This is without any time to death shift correction
Doing an 11 day shift
Another option is something inbetween. if we have a 90 day window instead of all previous values then it includes more recent variance.
or 60 days
Settled on last 90 days window
deaths falling faster than cases seems to be a good indicator that vaccine is working.
It would be see this for each region and each province.
cases / deaths.shift(avg_days_death)
is a rough estimate of daily CFR but a real one would have to track individual cases. But if detection rate doesn't change then it should show the effect of vaccines by lowering the CFR.Per province some death rate is so low it might be noisy so might need 21 day average.
We could also do the same for different ages per provice as we could collect that from the dashboard.