I’ve got time on my hands, and a pencil and some paper. So I thought I would take a look at the numbers being reported in the media. Presumably the smart people in the government have been doing this, in a much better way, for a couple of months.
--------
The number of confirmed positive tests is almost meaningless. Early on there were so few tests run the numbers were impossibly low. And even with solid testing there are many cracks in the system.
The testing was done on people who were symptomatic, meaning they
had probably become infected about 10 days earlier. The test is a lagging indicator and never measured the number of newly infected
But the data can be extrapolated because the number of dead is a solid statistic.
---------
About 440 (on average) people died each on March 28 and on 29.
If we assume a 1% fatality rate and 21 days from infection to death,
then that means 44,000 people became newly infected on March 7 and 8.
Looking back we see that 140 and 116 new infections were reported.
And those people were probably infected 10 days earlier.
So for every person that tested positive, about 320 untested people were newly infected. And they felt fine and began to spread the virus.
-------
About 890 people died each on March 31 and on April 1.
At 1% and 21 days,
that means 89,000 people became newly infected on March 10 and 11.
Looking back, 376 and 322 new infections were reported.
So for every positive test there were 255 newly infected people
that nobody knew about, who began to spread the virus.
-------
About 1165 people died each on April 2 and on April 3.
At 1% and 21 days,
that means 116,500 people became newly infected on March 12 and 13.
Looking back, 382 and 516 new infections were reported.
So for every positive test there were 233 newly infected people
that nobody knew about, who began to spread the virus.
-------
The number can also be teased out by using the positive test numbers.
There were limited tests and only highly symptomatic people were tested.
A decent guess is that everybody testing positive was infected 10 days earlier.
We can check the accuracy of the testing by comparing it to the number based on the death count — which is a solid number.
======================
Putting the number together by individual dates (no averaging):
Newly infected Newly infected
By Who Died By Positive Test
DATE 21 days later 10 days later
March 7 44,500 1,789
March 8 44,100 1,362
March 9 51,100 5,894
March 10 89,500 5,423
March 11 88,400 6,389
March 12 116,900 7,783
March 13 116,100 10,395
-----------------------—
The reported new infections is not close to the actual new infections.
Early on about 4% (1 in 25) infections were being caught, ten days later.
A week later it was up to 9% being caught, 10 days after infection.
A newly infected person has no symptoms, other than being around
infected people. So without massive testing of apparently healthy people lagging indicators are the best we’ve got.
Presumably the testing has been expanded to include asymptomatic people whose only indicator was contact with known infected people. That data is
not in the basic chart I used at the washington post.
www.washingtonpost.com/…
The most recent testing shows 30,000 new cases and 32,000 new cases.
Assuming they have loosened the testing access to include people with
minor symptoms, then the data lag may be only 7 or 8 days.
If they are off the 25 to 1 from early on (very low),
then 775,000 people were infected 8 days ago,
and 7,750 of them will die in 13 days.
If they are off the 11 to 1 from 3 weeks ago (probably low),
then 341,000 people were newly infected 8 days ago,
and 3,410 of them will die in 13 days.
If they are off by only 2 to 1, (probably high),
then 68,200 were newly infected 8 days ago,
and 682 of them will die in 13 days.
We will know in 13 days what percentage of people infected 8 days previously the test is catching. It is better than 4% but probably under 50%.
31,000 infections should mean 310 dead 21 days after infection (13 days after testing). If the number is still at 1,000 dead, then they caught only 31,000 out of 100,000 actually infected, or 31%.
Again, the smart people in the government should have known all this weeks to months ago. How long did it take before the best case of 100,000 dead was mentioned? How many weeks was that data hidden from the public? And what is the worst case projection the government came up with? Probably classified data because it would cause a panic or hurt Trump’s re-election chances
We will pass the hump without knowing it. The testing has big gaps in what percent of the newly infected are actually found, and that is for a point 8 days in the past. If the error is the same every day a decrease in reported cases might mean we hit the inflection point 8 days previously. But there are reports that hundreds of tests have been held up. Initially that would show as a drop in infection (which should show in the data) followed by a surge of infections when the tests are finally done (which should, and does, show in the data).
The more solid indicator for the inflection point in the infection rate will reliably show up in the death rate 21 days (approx) after it happens.
It is hard to guess today’s precise infection rate on past data because mitigating action (increased hand washing, increased disinfecting, social distancing, shelter in place, face masks, quarantine) should be lowering the rate, but by how much? And when one state goes into lock down but another state does not, how does that affect the rate? We will know 21 days after the fact.