Sunday, March 22, 2020

US: That "Hitch" is Gone

Yesterday's "hitch" in new cases doesn't pay out. 


as the US creeps still closer to Italy's path:


New cases in one day:

Italy    6,419  (14%)
US    6,982  (35%)
Canada    239  (22%)
China    45

US cases per capita are 9.2% of Italy's; a week ago we were at 2.2%.

But US cases per capita are now 2.1 times the same number for the rest of the world:


Finally, I should mention how I calculate my numbers. Instead of using WHO's data, which are given day-by-day for a full day, I use the current data that appears on Worldometer, owned by the software company Dadax. Since I don't look at the data every time the same day (but usually do so between 5-7 am, depending on when my cats want breakfast), I normalize the numbers to 24 hours. So if X new cases appeared in 25.5 hours, the number I report is (X/25.5)*24.

12 comments:

David in Cal said...

I have been using World-O-Meter as a data source. https://www.worldometers.info/coronavirus/#countries I checked it a few times and it agreed with WHO. I noticed something strange. Late yesterday, it showed a number of new US claims over 7000. There is a "Yesterday" button. Right now (8:04 PDT 3/22) it shows 2944 for NOW and 4824 for YESTERDAY. I hypothesize that this chart puts late claims into the current day, but later re-allocates some of them to the subsequent day.

As I recall, the re-allocated number for 2 days ago was around 5000. I'm not sure, but I think the figure 3 days ago was also in that range. And, as I said, the re-allocated number for yesterday is 4824. These figures would indicate that new claims have leveled off.

This sounds too good to be true. What am I missing?

Cheers
David

Ned said...
This comment has been removed by the author.
Ned said...

If you go to the USA country page, you can see the historical data (e.g., scroll down to the chart for "daily new cases" and hover over it). Here are recent days:

https://www.worldometers.info/coronavirus/country/us/

Mar 17: 1748
Mar 18: 2848
Mar 19: 4530
Mar 20: 5594
Mar 21: 4824

So yes, yesterday was down slightly. On the other hand, today's count is already at 6084 and there's still seven hours to go (they end each day at 00 GMT).

The same "noise" is visible in the daily deaths counts, too. The numbers for 3/20 and 3/21 were actually down slightly from 3/19 ... but today (3/22) is already at 86, which is basically right on the prior exponential function (using data from 3/8 to 3/20, the model would predict 85 new deaths today, compared to 86 currently).

Ned said...

Just for the heck of it, here are the exponential model's extrapolations for daily new deaths from Covid-19 in the USA, for the coming week:

Sun 3/22: 85
Mon 3/23: 110
Tue 3/24: 142
Wed 3/25: 183
Thu 3/26: 236
Fri 3/27: 305
Sat 3/28: 394

Disclaimer: those are simply naive extrapolations based on fitting an exponential growth model to the observations from 3/8 to 3/20 inclusive.

If one or two days are above or below the line, it's likely just noise. Hopefully there will start to be a reduction in the exponential growth rate, in which case we might see several days consistently coming in below those extrapolated values.

Ned said...

Yikes. Today's count of new cases just doubled. When I posted that previous comment half an hour ago, the new cases were at 6084. Now they're at 13,931.

David in Cal said...

Thanks, Ned. Do you know of any source for up to date figures on the number of Americans who were tested?

Cheers

David in Cal said...

That new count just jumped down to 8149.

Layzej said...

USians tested: https://www.cdc.gov/coronavirus/2019-ncov/cases-updates/testing-in-us.html

Number of specimens tested for SARS CoV-2 by CDC labs (N=4,524) and U.S. public health laboratories* (N=61,847)

Layzej said...

https://covidtracking.com/data/ shows 217,111

Layzej said...

Canada shows 54,956 at the national lab, and 58,165 tested by the provinces, but there is probably a lot of overlap because the national lab confirms provincial findings.

Ned said...

That new count just jumped down to 8149

That's much more in line with model predictions (~9000 new cases today). There's still 3 hours to go before they roll over to the next day, so I guess we'll have to wait and see what the final count is for today.

Layzej said...

54,956 tested by Canada's national lab is from earlier in the week. This is now showing 98,824 - https://www.canada.ca/en/public-health/services/diseases/2019-novel-coronavirus-infection.html#a2