Cont: 2019-nCoV / Corona virus Pt 3

[JeanTate]

Grim is the only suitable word

Make it 85k US deaths attributed to Covid-19 by Jan 1, 2021, and you have a deal.

Thanks! That works for me.

From a somewhat arbitrary 21 March, 2020, I calculated the daily increase in deaths from COVID-19, as a fraction of the previous day's total; per the WHO daily situation reports*. Up to 4 April.

The minimum is ~0.12; the maximum ~0.43.

Starting with 5854 COVID-19 deaths (cumulative) on 4 April, I estimate the soonest "85k US deaths attributed to Covid-19" will be 18 April, and the latest 28 April, with a mid (geometric mean) by 26 April. Allow ±1 day for all manner of variables.

One way to look at these numbers is this: if 85k is reached in May (or later!), it could be considered evidence that the massive/draconican measures put in place in March and April succeeded in "flattening the curve".

Sadly, I think the numbers tell us that 240k deaths is too optimistic (it is more likely than not that there will be more than 240k).

*there is no realistic number for 22 March, so I interpolated between 21 and 23 March.

 

[pipelineaudio]

He's an IT guy. Why should his opinion on a pandemic and how we respond to it be worth more than yours or mine?

When it comes to vaccines, you might want to look up exactly who it is you are talking about!

While I rail at him every time an update sends us back to the drawing board and our customers call for our heads, I'll never complain about his contribution to the health of every human on the planet

 

[Fast Eddie B]

Crescent's spreadsheet for USA coronavirus deaths by April 4 shows 7823. Actual number at 8:01pm shows 8407 so far. I say for the plus or minus that's pretty much spot on. Worse than expected though.

I was hoping to see that spreadsheet updated periodically with"actual" numbers posted alongside the "predicted" numbers.

 

[sir drinks-a-lot]

From a somewhat arbitrary 21 March, 2020, I calculated the daily increase in deaths from COVID-19, as a fraction of the previous day's total; per the WHO daily situation reports*. Up to 4 April.

The minimum is ~0.12; the maximum ~0.43.

Starting with 5854 COVID-19 deaths (cumulative) on 4 April, I estimate the soonest "85k US deaths attributed to Covid-19" will be 18 April, and the latest 28 April, with a mid (geometric mean) by 26 April. Allow ±1 day for all manner of variables.

One way to look at these numbers is this: if 85k is reached in May (or later!), it could be considered evidence that the massive/draconican measures put in place in March and April succeeded in "flattening the curve".

Sadly, I think the numbers tell us that 240k deaths is too optimistic (it is more likely than not that there will be more than 240k).

*there is no realistic number for 22 March, so I interpolated between 21 and 23 March.

Am I missing something or are you just projecting as if the increase in deaths per day will continue indefinitely as it has been?

 

[Doghouse Reilly]

Thanks! That works for me.

Now we need to discuss the stakes. Any proposals?

 

[Doghouse Reilly]

Am I missing something or are you just projecting as if the increase in deaths per day will continue indefinitely as it has been?

It seems that way. Based on her projections, I'll lose the bet in 2-3 weeks.

 

[Darat]

Expert enough in what? Managing a foundation? I'm talking about epidemiology.

I'm not saying he knows nothing about it, but that there are actual experts out there that I'd trust more to know about this thing.



What the hell does that have to do with whether he's an expert?

It answered the question you asked.

 

[JeanTate]

Great!

Over in FiveThirtyEight, founder Nate Silver gives a similar, longer explanation: Coronavirus Case Counts Are Meaningless*

The * is a sub-title: "Unless you know something about testing. And even then, it gets complicated."

Note to thread: SG has been shouting this since the beginning of the thread until she gave up bothering.

SG: my totally subjective assessment is that there are very few who post regularly to this thread (and its ancestors) who take the reported case counts at face value.

Reported deaths due to COVID-19? Perhaps some, even many, feel this is pretty reliable.

This, from WaPo, is sobering: "Coronavirus death toll: Americans are almost certainly dying of covid-19 but being left out of the official count" (link).

Maybe some reader who knows the details - not only for various parts of the US, but elsewhere in the world - could comment on how a death is attributed to COVID-19, and how the method/process differs from deaths attributed to cancer, say (I guess attributing a death to "gunshot" is more straight-forward, even if murder/accident/suicide is less so).

It's been noted, more than once, that the numbers of Wuhan/Hubei COVID-19 deaths are under-estimates. If only because in the early days there was often little to no testing.

I guess that where there have been few COVID-19 deaths - Singapore, say, or Taiwan - the reported numbers are unlikely to under-represent the actual numbers. Where resources are tight/stressed - as in New York these days - under-representing is certain.

 

[JeanTate]

Am I missing something or are you just projecting as if the increase in deaths per day will continue indefinitely as it has been?

(my hilite)

Yes.

And no.

26/27 March had the lowest growth in that period (~0.12), and 24/25 March the highest (~0.43; leaving out 19/20, which is 0.50).

Before obtaining the data, I had expected a fair bit of scatter at the beginning of the period, but not much towards the end. I was wrong. And while there is indeed a trend, a linear fit doesn't get to ~0.12 until ~17 April.

YMMV, of course, but my bet is that Dogghouse Reilly will lose the bet with Roboramma by the first week of May at the latest, and could lose it by the last weekend of April. I truly hope that I'm wrong.

 

[Doghouse Reilly]

(my hilite)

Yes.

And no.

26/27 March had the lowest growth in that period (~0.12), and 24/25 March the highest (~0.43; leaving out 19/20, which is 0.50).

Before obtaining the data, I had expected a fair bit of scatter at the beginning of the period, but not much towards the end. I was wrong. And while there is indeed a trend, a linear fit doesn't get to ~0.12 until ~17 April.

YMMV, of course, but my bet is that Dogghouse Reilly will lose the bet with Roboramma by the first week of May at the latest, and could lose it by the last weekend of April. I truly hope that I'm wrong.

I understand exponential growth, projections, etc., but to me it's just bizarro world thinking. This virus has been across the world for a few months now, and in the US for almost 2 months, yet the entire global death toll is under 85k, but somehow in less than a month, the US will surpass the entire global death toll?

If you turn out to be right, I will most certainly have been proven epically wrong. And if that comes to pass, I will certainly admit it and eat crow. But I think your projections are absurd.

 

[sir drinks-a-lot]

YMMV, of course, but my bet is that Dogghouse Reilly will lose the bet with Roboramma by the first week of May at the latest, and could lose it by the last weekend of April. I truly hope that I'm wrong.

Thanks for the clarification of your projection. I also think Doghouse Reilly will lose the bet, but much later in the year. Let's hope we're both wrong and Doghouse Reilly cruises to victory!

 

[Doghouse Reilly]

Thanks for the clarification of your projection. I also think Doghouse Reilly will lose the bet, but much later in the year. Let's hope we're both wrong and Doghouse Reilly cruises to victory!

This is great, everyone's rooting for me! If they don't, they're cruel and heartless sociopaths!

 

[Ryokan]

Boris Johnson in hospital because of corona.

 

[Samson]

Boris Johnson in hospital because of corona.

He must have acquired a high virus load.

 

[calebprime]

Thanks for the clarification of your projection. I also think Doghouse Reilly will lose the bet, but much later in the year. Let's hope we're both wrong and Doghouse Reilly cruises to victory!

agreed. I'd rather be wrong about my pessimism than right.

 

[Doghouse Reilly]

Boris Johnson in hospital because of corona.

Apparently for tests, not treatment, as he has persistent fever. He seemed to be doing pretty well in the video he posted yesterday, didn't appear short of breath or anything.

 

[crescent]

Crescent's spreadsheet for USA coronavirus deaths by April 4 shows 7823. Actual number at 8:01pm shows 8407 so far. I say for the plus or minus that's pretty much spot on. Worse than expected though.

I was hoping to see that spreadsheet updated periodically with"actual" numbers posted alongside the "predicted" numbers.

I'll try to update it tomorrow. I started yet another column on the same day that I first posted it, that new column starts with the end-of-day results for the first date, that seems like a bit of a better apples-to-apples comparison.

It is on my other computer though, down in the basement where I've been doing the work-from-home thing. Not much in the mood to spend time down there on my day off.

Going from memory though, I think we are still running at better than a 20% daily increase in American fatalities. We'll surpass the death toll from H1N1 swine flu tomorrow (Monday April 6) or the day after. We'll match the average annual death from influenza on Easter Monday, more or less.

Assuming we keep up at a 20% daily growth.

 

[Solitaire]

Going from memory though, I think we are still running at better than a 20%
daily increase in American fatalities. We'll surpass the death toll from H1N1
swine flu tomorrow (Monday April 6) or the day after. We'll match the average
annual death from influenza on Easter Monday, more or less.

Assuming we keep up at a 20% daily growth.

Since the US has been under quarantine regulations for the past two weeks,
I've been measuring regular rates of just a tiny bit under 15% daily growth.
So by 2020 April 15 we should see only 1.25 million cases in the United States
and 0.5 million cases in New York state.

I calculated the percentages from the last few numbers from an article in Vox
titled: Conformed Coronavirus Cases by Dylan Scott and Rani Molla.

 

[Minoosh]

India may have lucked out by catching a few early cases that popped up in remote areas. A plethora of religious sites and festivals for multiple faiths are slowly accelerating the cases (leading to condemnations of those followers in every direction).

It will be interesting (yes, also tragic) to see the spread play out in a country already under lock down before reaching a frightening pace.

Granting that 8,000 other variables weigh in, if a poor country with good discipline flattens their curve while a modern but obstinate country breaks its health care system, that would be quite a statement.

My *impression* of India is that there's so much infectious-disease background as "noise" that teasing out coronavirus illnesses/deaths will be very, very difficult, and that the PTB will take advantage of that to lowball the count.

Whereas with China, I'm confident the government knows exactly how many cases/deaths there have been, even if the numbers they publicly release are bogus.

 

[Fast Eddie B]

I'll try to update it tomorrow.

Thanks. Looking forward to it.