COVID-19 and New Hampshire

Recently, I was in a friend's Facebook comments (not recommended, I know) in a post about his disappointment with New Hampshire Governor Chris Sununu's decision to close the beaches. In these comments there was some notion that things are not bad here and COVID is not "exponentially bad", especially when "compared to New York".  So it pushed me to look for some modeling for the COVID-19 outbreak in New Hasmphire, which is harder to find than I thought.  I did eventually find some stuff to show that NH is definitely seeing cases grow exponentially, but the data was not complete.  So today, I spent an hour going through the data and now have a spreadsheet where I am tracking the outbreak for NH and modeling in a very simple way against a theoretical exponential model.  The raw data spreadsheet can be found at the bottom of this post.

I want to use this space to share the charts I made and explain them, and then going forward, the google sheet at the bottom of the page will be able to be accessed to see progress.

Here are a few things to know.  Data is messy, particularly when making simple models and simple projections, to help with this, I am going to fit the model to a 5-day average. This will help account for fluctuations between days where people might be more likely to get tested, or tests might come back faster due to any number of factors like proximity to a weekend of a holiday. However, when I show you the growth curves, they will be on the actual total number of cases.  Everything in this post should be able to be understood if you have taken introductory statistics.


In this chart, the blue line is the total number of confirmed cases in New Hampshire (479 at the end of April 2) and the red line is an exponential growth curve.  This growth is a new cases increase of 18.5% each day, meaning that the number of new cases and total cases will double every 4 days.

Source: Data Gathered from: NHPR Live COVID-19 Blog 

When I look at the last 2 weeks of data, my 5 day average has an average growth of 15.5% with a standard deviation of 0.08%.  This means the model estimate of 18.5% is feasible but possibly over predicting a tad.  Thus, I will use the model above AND change the model to account for 15.5% growth to show two models of what this continued growth would look like over two weeks.

Note: The header here says 18% but it should be 15.5%.  Apologies for not editing it.

For each of these models (note the blue line still exists in both, but it now very hard to see), this assumes sustained growth, which is the exact thing we are trying to stop.  The difference between 18% and 15% growth is 1000 cases over the next two weeks, a big difference. The state-wide stay at home order is 7 days old, so we still have one more week (of half of the two predictions above) before we would expect to see this growth slow.


Finally, the last graph I will share is one of new cases.  This is the one where we will more obviously see sustained growth slowing.  This is what that currently looks like:

In this case, our data is 2 days behind because we are modeling according to a 5-day average, so the most recent 5-day average is March 31 (which is the average of March 29 through April 2). This looks the same as above because the growth is the same, but when the growth slows the shape will start to take a more bell-shape and will be easier to see improvement.



Raw Data: https://tinyurl.com/covid19nh