COVID-19 June 06, 2020

I've been showing the Gompertz curve along with both the infection data and the death data.  The Gompertz model failed long ago for the infections, which have plateaued for weeks rather than decreasing as the model predicted.  In the last couple of weeks, the death statistics have also departed from the Gompertz model.  Neither of these is surprising.

Today, I attempted to fit new curves to the data, ignoring earlier data, and focusing instead on trying to better match more current data.  That exercise failed completely for the infection data.  There is no Gompertz curve (or similar so-called logistic/S curves that will fit the data).  The death data is a different story, however.  If you focus only on the data from about mid- to late-April and ignore prior data, there is a curve that shows a reasonably good fit. It's a terrible fit with the earlier data.  You can see that misfit when I plot the curve and data on the old semi-log plots.  The curve and the data are very far apart right up to April.

The new curve suggests a total death tally of 120,000.  Note that this number is the number based on the data set that I'm using, which tends to be lower than others by 10-20%.  So, if you prefer other data sets, my 120,000 would probably translate to something like 130,000-150,000.  As always, this assumes the data continues to follow the curve.  There is no reason to suspect that it will.  What the new Gompertz curve does do is give a more current baseline that can be used to provide context for the data.  Are we staying on the curve?  Rising above?  Falling below? 

Also, I've updated stats for Texas, which I'm showing below.  Texas is one of the states that seemed to be showing a noticeable increase in infections.  The analysis seems to bear that out.  There are others that seem to be showing similar trends.  I don't have time to analyze all those data.  There are other websites that allow you to see stats by state if you are inclined to dig further.