The Modeling

I’ve added this section, which I’ll leave on the report permanently. Note that I’ve been (and continue) to report on and model “Reported Cases”. These reports come from the various states and I retrieve my data from www.infection2020.com and www.covidtracking.com. I regularly check their numbers against several other reporting sources (including the CDC) to make sure the numbers track well. To be reported as a case, the subject must have been tested at the state level, with a positive finding. It is not necessary that the case be validated by the CDC, as was the case in early March. Some jurisdictions are now reporting likely cases that have not tested positive. As of April 17th, states report likely, presumed, and confirmed cases all as confirmed.

My model of cumulative reported cases was developed by fitting a Logistic functional form to the reported data. I calculate new parameters for the model from time to time. I translate Cumulative Cases to Active Cases by assuming the time from case reporting to case recovery is 7 days. I hypothesize this by assuming a 13-day disease cycle, but 3 days from symptoms to sample collection, 3 days from sample collection to test reporting, and another 7 days to case resolution on average. Technically, these are “estimated reported and still active cases”, but I’ll often refer to them in shorthand as simply “active cases” or “active reported cases”.

Reported Cases does not encompass everyone who has contracted the disease in U.S. society. Some do not seek testing, some in the early days couldn’t get a test (not so true anymore), and some contracted the disease, but were asymptomatic. We cannot know, and may never know, how many citizens actually have the infection. That said, so long as the protocol for testing and reporting remains stable, the data should be stable and represent the course of the disease. I have not modeled mortality at this time, as I believe others are doing this well.