CDC revises fatality rate
-
wrote on 25 May 2020, 19:58 last edited by
Of course. The infection rate of the 21k numerator is 100%. Those are the Covid deaths.
-
I dont see either one of your points. - no conceivable demographic details get you from a 0.25% population fatality rate to a 0.26% infection fatality rate with a 20% serology result. And no conceivable fact about nursing homes or inter-state comparisons gets you there either.
wrote on 25 May 2020, 19:59 last edited by@jon-nyc said in CDC revises fatality rate:
I dont see either one of your points. - no conceivable demographic details get you from a 0.25% population fatality rate to a 0.26% infection fatality rate with a 20% serology result. And no conceivable fact about nursing homes or inter-state comparisons gets you there either.
It seems conceivable that the 20% underestimates the rate of infection of the pool of folk who comprised the numerator of the fatality rate. I don't mean to make the tautology that if you died of it then you had it, i mean to say that they came from an identifiable cohort (nursing homes?) with far greater than 20% infection rate.
-
wrote on 25 May 2020, 20:26 last edited by
Right. So sample bias in serology because serology test recipients were unlikely to be nursing home residents.
That’s a point, though even if every New Yorker over 75 was positive and unaccounted for in the serology sample that would bring us to 25% infection rate rather than 20%
-
Right. So sample bias in serology because serology test recipients were unlikely to be nursing home residents.
That’s a point, though even if every New Yorker over 75 was positive and unaccounted for in the serology sample that would bring us to 25% infection rate rather than 20%
wrote on 25 May 2020, 20:48 last edited by LokiNYC deaths by age per100,000
Over 75 is greater than all the other categories combined by well (vastly) more than double.
https://www.statista.com/statistics/1109867/coronavirus-death-rates-by-age-new-york-city/
-
wrote on 25 May 2020, 20:55 last edited by
In NYC you were almost 100 times as likely to die if you were over 75 than under 44.
-
wrote on 25 May 2020, 21:04 last edited by
@Loki said in CDC revises fatality rate:
NYC deaths by age per100,000
Over 75 is greater than all the other categories combined by well (vastly) more than double.
https://www.statista.com/statistics/1109867/coronavirus-death-rates-by-age-new-york-city/
That link implies 16500 total deaths in nyc rather than 21000. (196/100000)*8400000 = 16500
-
wrote on 25 May 2020, 21:17 last edited by
@Loki said in CDC revises fatality rate:
In NYC you were almost 100 times as likely to die if you were over 75 than under 44.
Loki by now we all fully understand your discount function on Covid deaths. Does it have any bearing on the accuracy of the CDC model?
-
wrote on 25 May 2020, 21:21 last edited by
@Loki said in CDC revises fatality rate:
In NYC you were almost 100 times as likely to die if you were over 75 than under 44.
This sort of distinction seems important, in a debate about whether to shut down a society. And all the biggest impact rhetoric of the debate, such as counts of lives lost or lives that could have been saved, completely ignores it.
-
wrote on 25 May 2020, 21:27 last edited by jon-nyc
If we were talking about lockdown measures I wouldn’t have found the comment out of the ordinary.
But that makes some sense out of the inability to see the obvious arithmetic impossibility of the CDC estimate in NY. I thought we were arguing about a CDC model not lockdown measures. But I guess we’re always arguing about the lockdown.
-
@Loki said in CDC revises fatality rate:
In NYC you were almost 100 times as likely to die if you were over 75 than under 44.
Loki by now we all fully understand your discount function on Covid deaths. Does it have any bearing on the accuracy of the CDC model?
wrote on 25 May 2020, 21:39 last edited byIt is not impossible that a model built from a large set of data will seem arithmetically at odds with some subset of that data, which might be an outlier. Are we concentrating on NYC because it seems to be an outlier, while ignoring other sets of data which seem to corroborate the model? The CDC model is actually under no obligation to conform to every subset of the data, it is meant to predict in general. And yes, it is conceivable that the NYC numbers imply fatality rates which overestimate the general case.
-
wrote on 25 May 2020, 21:41 last edited by
“My model of heat dissipation in ceramic tiles was confirmed by 134 out of 135 Space Shuttle missions.”
-
wrote on 25 May 2020, 21:44 last edited by
I get the model can’t conform to every conceivable subset of data but your biggest outbreak by far isn’t just another subset of data.
If the model is going to have any useful predictive power it can’t miss the big important cases.
-
wrote on 25 May 2020, 21:49 last edited by
I wonder if the CDC has a better handle on the NYC numbers than we do in this thread. You appear to have been off in your total death count by about 25%, for instance.
-
wrote on 25 May 2020, 21:53 last edited by
No. The NYDOH counts 21k. 4700 didn’t get a pcr test because they died at home.
See the first hand story of the EMS guy that George posted in early April.
-
wrote on 25 May 2020, 21:59 last edited by
Ok then. For the record, I am betting that the CDC model will be a better predictor of future national numbers than the NYC numbers will be. I could certainly be wrong. I assume your bet is the opposite?
-
wrote on 25 May 2020, 22:02 last edited by
CDC is using the higher NY number
I can’t upload the screen shot because it’s too big
-
wrote on 25 May 2020, 22:02 last edited by
No I think they’re wrong by minimum a factor of 2.
My guess is 0.5<IFR<0.75
-
wrote on 25 May 2020, 22:18 last edited by
I assume the fatality rate will inevitably decrease after the first wave of infections, since those who were at the bleeding edge of risk will have already died or become immune.
-
wrote on 13 Jul 2020, 16:58 last edited by jon-nyc
@jon-nyc said in CDC revises fatality rate:
No I think they’re wrong by minimum a factor of 2.
My guess is 0.5<IFR<0.75
CDC revised the estimate again. As a reminder, in the first post of this thread, their 'best estimate' scenario had an IFR of 0.25. Their new update, published Friday, increased it to 0.65%, smack in the middle of my range.
https://www.cdc.gov/coronavirus/2019-ncov/hcp/planning-scenarios.html
-
@jon-nyc said in CDC revises fatality rate:
No I think they’re wrong by minimum a factor of 2.
My guess is 0.5<IFR<0.75
CDC revised the estimate again. As a reminder, in the first post of this thread, their 'best estimate' scenario had an IFR of 0.25. Their new update, published Friday, increased it to 0.65%, smack in the middle of my range.
https://www.cdc.gov/coronavirus/2019-ncov/hcp/planning-scenarios.html
wrote on 13 Jul 2020, 17:01 last edited by@jon-nyc said in CDC revises fatality rate:
@jon-nyc said in CDC revises fatality rate:
My guess is 0.5<IFR<0.75
CDC revised the estimate again. As a reminder, in the first post of this thread, their 'best estimate' scenario had an IFR of 0.4. Their new update, published Friday, increased it to 0.65%, smack in the middle of my range.
https://www.cdc.gov/coronavirus/2019-ncov/hcp/planning-scenarios.html
I’ll take credit too as the Diamond Princess example I was using very early on seems to have stood the test of time.