| Bofferbrauer2 said: Luxembourg now gonna start reall mass testing. From Monday through end of Juli the country wants to test 20000 persons per day, or 3% of it's population each and every day. That way the country wants to ensure that there are no unknown asymptomatic patients spreading the disease and to know exactly who got the disease who didn't. In any case the disease in slowly going away. On Thursday, out of 750 tests only one had been positive. Luxembourg also has expanded on the statistics they release, now including R0 rates and how they are evolving. https://msan.gouvernement.lu/en/graphiques-evolution.html#sg Also comes with a handy description how R0 exactly works: https://msan.gouvernement.lu/dam-assets/covid-19/graph/Description-reproduction-numbers.pdf |
Interesting, and good job by Luxembourg! Hopefully with that early warning system in place you can prevent a second wave, keep the water calm.
I've probably been calculating Rt eff wrong.
I've been comparing week over week changes, then take the 7Root() from the change in reported cases (average of 3 days, compared to avg of 3 days exactly one week ago) to get the avg daily change factor without the weekly cycle in reporting interfering. Then I take that number to the power 5.1 (the median incubation period) to reflect how many people each infected person infects.
I guess it's the same thing in a roundabout way.
For example when I apply that to Luxembourg
23 cases reported from May 20 to 22
29 cases reported from May 13 to 15
Factor 0.7931 for 7 day comparison, 0.967x per day (7Root), Rt eff 0.845 (0.967 to the power 5.1) for May 21st
The graph shows 0.85 for the 20th, I guess it comes out the same :)
Looking at changes in reported cases really needs more cases to stay accurate. Below 100, fluctuations in reporting start to amplify a lot.
I've also been redistributing outliers here and there when a country I'm tracking suddenly throws up a whole bunch of old cases. For example France is really bad with outliers, -141 cases one day, spike of 3640 cases earlier. Plus a couple days ago France resurrected 217 people. The UK also had -525 cases a few days ago. Ireland also put up a nice outlying spike not long ago. I mostly keep the messy data as is in the primary graphs (except the negative deaths), but need some adjusting for the growth comparisons since it will skew the stats twice for several days. Kalman filtering by hand...
