I have been attempting to understand the nature of the “reproduction number” in epidemiology and to this end I’ve enlisted the help of the R0 package for R.
For a good overview of what the reproduction number is and why it is so important please check out this article.
This package allows you to estimate R0 given a time series of case counts. I have used the data from the 2019 Novel Coronavirus COVID-19 (2019-nCoV) Data Repository in order to perform these calculations.
Further details including R code are supplied in my GitHub repo, but assuming it behaves like SARS in terms of generation time (which is a big assumption on my part) then we can generate the following plots.
Here are the predictions for case counts using the various different simulation methods available in R0:
If we use the TD or “time-dependent” method (see here again) then we can determine effective R over time:
R seems to be tending towards some value between 2 and 3.
It’s important to note a couple of things here.
- The reproduction number plotted here is not an intrinsic property of the virus. It is a feature of the virus and the population plus the measures taken to combat it. It will vary over time and indeed over space. R0 (the number, not the package) is the intrinsic reproduction number.
- I am not an epidemiologist and I have made a number of half-educated guesses. Do not take these numbers as anything more than an experiment in using the R0 package.
The full R-generated report is here.