I have no idea what you mean. The entire point is that confirmed case data does not give us a good guide on true infection numbers, but we do need to include a lag because deaths don't happen the day after infection and are not reported for some time after that
-
-
Replying to @GidMK
What I mean is that the 83M total infections is February–December 2020, but you are using total deaths up to Feb 22. Even with lag, you are adding deaths in Jan, but not infections. From Dec 30 to Jan 30 there is 35% more cases so infections is higher than 83M. 25% higher?
1 reply 0 retweets 0 likes -
Replying to @JonStanich
This is the challenge. It is plausible that you could use the deaths number for mid-late Jan to calculate IFR, because infections up until 31/12 would be recorded as deaths at the earliest on 15-22/01
2 replies 0 retweets 1 like -
Replying to @GidMK @JonStanich
But given the reporting lag, people who got sick on 31/12 would mostly not appear in the official figures until the start of Feb, and you wouldn't see most of them in the deaths data until about now
1 reply 0 retweets 1 like -
Replying to @GidMK @JonStanich
So if we use the data from mid-Jan, we would get an IFR of 0.5% (410k deaths), but the true value lies somewhere between 0.5-0.6% as I said
1 reply 0 retweets 1 like -
Replying to @GidMK
Why do you not account for additional infections from Dec 30 to Jan 15? Cases went up 15% in that time, so total infections must have gone up also.
1 reply 0 retweets 0 likes -
Replying to @JonStanich
I don't think you are understanding how infections and deaths work. People rarely die immediately on being infected, that's the whole point of a lag. Deaths are also almost never reported on the day that they occur
2 replies 0 retweets 2 likes -
Replying to @GidMK
I understand clearly. If you assume 400K deaths by Jan 15, and 83.1M infections by Dec 30 (assuming some of these infections lead to deaths by Jan 15) you get 0.48% IFR. That is more accurate than 0.6%, but still an estimate adjusted for lag & infections only up to Dec.pic.twitter.com/4ojnMtO1SU
1 reply 0 retweets 0 likes -
Replying to @JonStanich
So that is the absolute lower bound of your estimate, given the lag between infection->death. But as we discuss in the paper, reporting lags often take 2+ weeks, so a death that OCCURS on 15/01 doesn't appear on your graph until 30/1 OR LATER
1 reply 0 retweets 2 likes -
It actually gets even more complex than that, but this is the basics. Read the paper if you want the detail, there's a really complex inference process to fully understand the complexity here
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.