Chronology Current Month Current Thread Current Date [Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

# Re: [Phys-L] Fwd: Re: six more days in Okla

Hmmmmm: your best fit of cumulative cases vs time puts most datapoints below the red line.
Have you been best-fitting log cases vs time by any chance?
If so, you did what colleges teach - giving undue weight to HIGH values and all but ignoring low values. But I am pleased to see someone is interested in tracking the time series.

Brian W

On 4/1/2020 3:05 AM, bernard cleyet wrote:
Here’s mine: http://cleyet.org/covid-19/USA/Oklahoma/.

bc. obtains approximately the same result, as brian for the same days even though his data is different.

Both the deaths and the tested cases rates decrease 28, 29, and 30 March. Must be patient.

On 2020/Mar/31, at 21:09, brian whatcott <betwys1@sbcglobal.net> wrote:

Just cannot get a credible Exponential out of the last six days or so. High R^2 linear fit. Who'da thot! [not yet April, either! ]

https://imgur.com/a/RpbJMnq <https://imgur.com/a/RpbJMnq>

Brian W

On 3/30/2020 1:49 AM, brian whatcott wrote:
An exponential model makes fairly good agreement with the official data for total cases.
When a non linear regression app makes the fit
fall below R^2 = 0.95, a logistics model (of the kind used to fit the HIV epidemic well) starts to show a decline in "carrying capacity" i.e the point at which all the people who can be infected ARE infected and the exponential growth smooths out to a flat top, showing that high point falling from many millions to a relatively smaller number in the last few days for Oklahoma.(population 3M+)
It takes several days of below expected exponential growth for a curve fitter to invoke a smaller value for the "carrying capacity"in a logistic model - and that's what we have today - but it is likely just a pot-hole in the progression. If the shortfall in growth continues long enough, one will gain some confidence in predicting the top. Modelers who are paid to work in this area prefer the NEW cases per day, rather than the total cases. This has the usual problem of working from a change in numbers rather than the total numbers - it is very noisy, so that these folk use weekly new cases to smooth out their pot-holes. The graphs here use a more sensitive, but not very noisy, daily data method.

Brian W

On 3/29/2020 9:10 PM, Bill Norwood via Phys-l wrote:
Hi Brian Whatcott,
- Sorry, I’m dense on this one.
- I need something I can share with relatives.
- Cyp explain more verbosely.
Many thanks!
Bill Norwood

Sent from my iPhone

On Mar 29, 2020, at 6:41 PM, brian whatcott <betwys1@sbcglobal.net <mailto:betwys1@sbcglobal.net>> wrote:

On 3/29/2020 2:24 AM, bernard cleyet wrote:
and no posts. Have we all succumbed to the corona?

bc soon worried.
_______________________________________________
Forum for Physics Educators
Phys-l@mail.phys-l.org <mailto:Phys-l@mail.phys-l.org>
http://www.phys-l.org/mailman/listinfo/phy <http://www.phys-l.org/mailman/listinfo/phy>
I pinned up the local (Oklahoma) statistics for cases with time.
My logistic plot saw a downturn in the rate of infection for three days, and so it revised its peak population infected line from millions to ~ a thousand. It SO over reacts to good news!
https://imgur.com/a/LjNDR7Y <https://imgur.com/a/LjNDR7Y>
_______________________________________________
Forum for Physics Educators
Phys-l@mail.phys-l.org
http://www.phys-l.org/mailman/listinfo/phys-l