Statistical Modelling During the COVID-19 Pandemic

Statistical modelling has been trotted out and used like a show pony by politicians and public health officials around the globe during the current pandemic. Despite the fact that, time and time again, the modelling has proven to be completely inaccurate, governments have used it as a very sharp tool to instil fear into their citizens.  In this very brief posting, I want to look at a recent example that appeared on the Centers for Disease Control and Prevention (CDC) website.

 

Here is the CDC's most recent national forecasts for reported new COVID-19 hospitalizations per day for the month of December 2020:

 

Over the month of December 2020, eight Of the eight models, four national forecasts predict a likely increase in new hospitalizations over the next three weeks, four forecasts are uncertain about the trend or predict stable numbers and one forecast predicts a likely decrease in new hospitalizations.  Let's look at one sample day; for December 28, 2020, the number of new COVID-19 hospitalizations ranges from 2,300 per day to 23,000 per day, an order of magnitude higher.

In case you are interested, there are also state-level models as well.  Let's look at four examples:

 

1.) New York:

 

2.) California:

 

3.) Texas:

 

4.) Florida:

 

 

In all four cases, there is a very wide range of predicted new COVID-19 hospitalizations.

 

Let's close with this thought; when your local, state or national politicians trot out the "modelling pony", remember that statistical modelling is little more than a best guess and that we should not be alarmed or fearful when we hear that there "could be" very high numbers of hospitalizations over the next few weeks.  Only time will tell us which model is correct but there is one thing that we can count on; statistical modelling is far, far from scientifically sound.