COVID-19 Time Series Analysis – 3/18/2020

One of the things I’m seeing a lot of discussion around is “flattening the curve” and the effects of social isolation. I’m very curious as to why some countries/regions have seen very different growth patterns in confirmed COVID-19 cases. One red herring I’ve seen a lot is the comparison of the first few days/weeks of the Italy outbreak with the first few days/weeks of the US outbreak. The reason this isn’t an honest comparison, of course, is that with an exponential curve one cannot know up front what the value of the exponent is or whether that power curve pattern will continue.

Here’s some time series analysis of the outbreaks in the top regions, first by the world then by US states.

It’s very interesting to see the differences. Assuming the data collection is accurate, it would appear that China and S. Korea have inverted their curves (looks more like a sigmoid curve now) — but for how long? Italy, Spain, and France still seem to be in their initial power curve phase, whereas Iran seems to be increasing linearly (yet another sign that their data is bad, most likely).

In the US chart I also included the Diamond Princess as a point of interest. Note how in a very small, contained sample the number of cases went flat quickly. This may point to the great value of the extreme social isolation that some countries have imposed (from reports I’ve read, Italy seems to have failed this step).

Due to the small size of the images embedded in this WordPress blog, I’m also starting to add PDF versions afterwards that can be downloaded in case you want to see higher resolution.

Time Series analysis for Top Outbreaks in the World. Data from JHU COVID-19 project.
Time Series analysis for US States AND the Diamond Princess cruise liner. Note how the Diamond Princess quickly went flat.

Leave a Reply

Your email address will not be published. Required fields are marked *