How to visualize hyperlocal COVID data

Don’t let the third wave get you unprepared

Italy’s multilevel governance system is complex. We have national, regional, provincial, and communal levels. The most granular COVID data available in Italy are at the provincial level. After some reflection on how to better visualize COVID in Italian provinces, I came up with a solution. Ladies and gentlemen, please welcome the: log-new daily cases with loess smoother.

This ggplot2 chart uses the y-axis with a logarithmic (base 10) transformation. This is different from most of the things you will see around. First, the logarithmic scales are (in theory) more suited for cumulative data. Yet, they have huge advantages in this context. They allow us to zoom on the lower numbers and to avoid enthusiasm when the situation changes. In fact, as the numbers grow higher, the curves (visually) move less, the same happening when numbers decline.

The smoother I have chosen is the loess, which is ggplot2's default. Given its statistical properties, this smoother (with this particular kind of data) is perfect, showing us the overall trends, reducing the noise from changes in the time series. The alternative would have been the following chart: