Fixing Coronavirus viz with actionable metrics
We’re doing the Coronavirus visualisations all wrong, claims Rémy David. Instead of mapping the changes to the increase or decrease of the virus (which you’d typically get in the now-familiar logarithmic graphs), we should be measuring the rate of change of the change.
Before your mind melts, David explains it thus:
If we take a car analogy, the total number of cases is the car mileage, the daily number of cases is the car speed, the daily number of cases variation is the car acceleration (positive if the car is accelerating, negative if the car is breaking) and the daily number of cases rate of variation is the way the car is accelerating: is it accelerating more, or is its acceleration slowing down (its speed still increases, but at a slower pace). In the case of the outbreak, we are mostly interested in the way the car is accelerating, right? So why is it that everybody is showing us the car mileage?Rémy David, Towards Data Science
By using linear interpolation, you end up with a straight line that has either a positive or negative inclination, and is pretty easy to read. Down good, up bad. Once it crosses zero, you’ve passed the peak of the outbreak.
Rémy does note that he’s not an epidemiologist, and is basing his graphs on pure math, discounting the myriad factors that the pros use. It’s still a powerful way to get to the “actionable metrics”, and is much easier to read than its logarithmic cousin.