Re: [Phys-L] void problem for 10 days - more landfill

[brian whatcott <betwys1@sbcglobal.net>]

On 5/4/2020 1:00 AM, bernard cleyet wrote:
> So:
>
> Tiny Andorra has crushed their local epidemic:   http://cleyet.org/covid-19/Europe/Andorra/
>
>
> My current residence doing much better than a previous:  http://cleyet.org/covid-19/USA/California/Monterey%20and%20Santa%20Barbara%20counties%20compared/
>
>
> bc …. filling the void
> _______________________________________________
> Forum for Physics Educators
> Phys-l@mail.phys-l.org
> http://www.phys-l.org/mailman/listinfo/phys-l
bc reminds me here that sigmoid plots can serve well.
Sadly I find that forcing cumulative data into a tidy bell-shape curve 
or even a skew curve of the kind that the math lion of Georgetown helped 
me with, by differentiating its fitting function in order to describe 
its rate, can depart from ground-truth.

In this situation, I shrugged my shoulders and applied a best running 
10-day linear fit to the aggregated data, and plotted these rates daily 
to form the OK & US Case and Death Daily Rates, using best fits to a 
four-term polynomial which obviously has little to no predictive value. 
One observation emerges that is interesting: a weekly cycle.

https://imgur.com/YfsUnd0 ; OK Daily Case Rates to May 3

https://imgur.com/6WD3HdT ; OK Daily Death Rates to May 3

(It is notable that the OK Governor asserted that he will reconsider his 
decision to stage a reopening if the Hospital Admissions data (a 
surrogate for CASE rate?) trend up
again, so I may be witnessing the death of good intentions)

https://imgur.com/jBnvANy ; US Daily Case Rates to May 3

https://imgur.com/8AaVksf ; US Daily Death Rates to May 3

There is a voice of caution which says plotting rates
 in this way amounts to averaging of averages -
 a known defect of data modeling - but I try not to listen!

Brian W