Early, but possibly very good news.
https://www.captainsjournal.com/2020...es-in-america/
The Current Trajectory Of Confirmed Covid-19 Cases In America
BY HERSCHEL SMITH
14 hours, 40 minutes ago
In my ongoing coverage and analysis of Covid-19 in America (updated almost daily), I included a much earlier graph with a curve fit, at the time, exponential and with a very high correlation coefficient. The graph from 3/23 looked like this.
The doubling time was computed as:
ln(2) / 0.2988 = 2.32 days
The graph has been unsettled lately, until last night and today. I have received requests to update the curve. I said I would have to jettison the exponential curve fit and go with a polynomial (see original post), and today I did that. The exponential model was massively over-predicting cases going forward and the correlation coefficient had begun to degrade. The revised curve is below.
There is a remarkable difference. The doubling time depends on where you are on the graph. It’s a third-order polynomial. Currently, the doubling time is 4.1 days, versus the value of 2.32 days computed not too many days ago. The correlation coefficient is very high, and the curve is stable and well-behaved.
Here I am not weighing in on or performing analysis of the reasons for this. There could be many, or only one, or some combination of causes. Some readers may posit “social distancing,” others may point out that the testing rate has change because slightly symptomatic patients are not being tested, others may postulate that herd immunity may be playing a factor (i.e., it’s possible that many millions of Americans have already been exposed to and infected with the virus and had little to no problem with it), and still others may postulate that PPEs, hygiene protocol and the reluctance to go to hospitals may be playing a role (my own daughter, a surgical NP and first assist who also has to spend copious time in the ER) observes that numbers of patients entering hospital care is down.
Again, I am making no claim whatsoever as to reasons for this. I am only mathematically modeling this phenomenon, and I can conclusively say that there is a remarkable difference between doubling time and trajectory today and a week ago.
UPDATE:
Per request, this is a picture of the previous exponential fit versus the polynomial fit. It’s QAD (quick and dirty), with no bells and whistles.