Wouldn’t it make sense to measure this based on number of deaths / (number of deaths + recoveries)? The media currently reports this stat as deaths / total cases, but given how quickly the virus is spreading this seems to be an incorrect method. The result is material (~18% in the former vs. ~2% in the latter).
Bruh why ru asking an anonymous finance forum as if anyone here remembers anything past AP Bio
I agree with you I would definitely measure it as you said number of deaths / (number of deaths + recoveries). I did not realise it would be that high 18% is crazy!
No, deaths/(deaths + recoveries) would be a bad estimate since recoveries take longer than deaths. For a ~ exponential function, this effect is really large, and you will overestimate by a lot. Corollary: the method that the media is using: deaths/total would be an underestimate since deaths don't occur instantly. (there are other factors like unreported cases as well, but we are not accounting for those here) A alternative (arguably better) estimate of mortality rate would be first estimating the expected time from infection to death (from those that died, call this value x); using deaths/(total as of x days ago). Of course, this method suffers from a smaller sample size when you estimate x, but you might consider this "unbiased" despite having higher variance (from smaller sample size, read: bias-variance trade-off)
Exactly this, a cohort based approach