Some potentially useful numbers I’ve been working on estimating:
1. The number of days lag between registered cases and deaths
2. The adjusted CFR for each country taking this into account
The method is essentially to try different lags (dividing current deaths by cases from x days ago) and see which length of lag gives a constant CFR over time (normally CFR increases with time as the growth rate of cases slows earlier than that of deaths).
Here are the results for a few countries:
China: 9 day lag, CFR=4%
USA: 7 day lag, CFR=6.5%
Italy: 4 day lag, CFR=14.5%
Spain: 2 day lag, CFR=10.5%
Germany: 10 day lag, CFR=3.5%
France: 10 day lag, CFR=24%
Switzerland: 6 day lag, CFR=4%
UK: 4 day lag, CFR=18.5%
I’m not sure about these, especially UK but they do create nice constant values for CFR over a period of 2-4 weeks (UK only 10 days) which suggests a predictable pattern, despite variation in testing.
The France result is also not quite as consistent as the others and is surprisingly high so I don’t quite trust it either. I could make a case that for an estimate of 7 day lag and 20% CFR.
Some potentially useful numbers I’ve been working on estimating:
1. The number of days lag between registered cases and deaths
2. The adjusted CFR for each country taking this into account
The method is essentially to try different lags (dividing current deaths by cases from x days ago) and see which length of lag gives a constant CFR over time (normally CFR increases with time as the growth rate of cases slows earlier than that of deaths).
Here are the results for a few countries:
China: 9 day lag, CFR=4%
USA: 7 day lag, CFR=6.5%
Italy: 4 day lag, CFR=14.5%
Spain: 2 day lag, CFR=10.5%
Germany: 10 day lag, CFR=3.5%
France: 10 day lag, CFR=24%
Switzerland: 6 day lag, CFR=4%
UK: 4 day lag, CFR=18.5%
I’m not sure about these, especially UK but they do create nice constant values for CFR over a period of 2-4 weeks (UK only 10 days) which suggests a predictable pattern, despite variation in testing.
The France result is also not quite as consistent as the others and is surprisingly high so I don’t quite trust it either. I could make a case that for an estimate of 7 day lag and 20% CFR.