Read on Twitter
I& #39;ve refrained from making COVID-19 charts up untill now. But the current situation in Belgium is dramatic, and now is a good time to show the power of a lesser known chart type 
https://abs.twimg.com/emoji/v2/... draggable="false" alt="🧵" title="Thread" aria-label="Emoji: Thread"> https://twitter.com/maartenzam/status/1319063929994252288">https://twitter.com/maartenza... 1/12
Let& #39;s start simple. These are the daily mortality and hospitalisation numbers for Belgium. We are all too familiar with this kind of chart 2/12
We are now also familiar with the so called "weekend effect", which is the reason why most reporting today is done using weekly averages. These makes it a little easier to separate the signal from the noise 3/12
All interesting, but let& #39;s remove time from the x axis to better see the link between hospitalisations and deaths. We put the number of hospitalisations on the x axis, the number of COVID-19 deaths on the y axis, and than we connect the dots according to time 4/12
Not really convincing yet :) The reason? You& #39;ve definitely heard about this one too: the exponential character of the spread of the virus. We should use logarithmic scales to get a better view 5/12
Let& #39;s add the weekly averages again to see the trends 6/12
So, what are we looking at here? There is a clear loop in the data: somewhere end March-beginning of April the number of hospitalisations peaks (curve starts moving to the left). A little later, the number of deaths peak (curve moves down again) 7/12
You can also see the hysteresis in the numbers: at the same number of hospitalisations we have relatively low number of deaths when we move up the chart, and higher number of deaths when we move down https://en.wikipedia.org/wiki/Hysteresis ">https://en.wikipedia.org/wiki/Hyst... 8/12
There is also a smaller loop in August: back then we managed to control a flare-up of the epidemic. Lesson learned: act quickly, when numbers are still low 9/12
So, where are we today? We are right back in the situation of mid March, when we went in complete lockdown. I find this really astonishing 10/12
A glimpse of hope: the latest numbers hint at a lower mortality. But because of the delay in the reporting of deaths, the last arrow on the curve will probably start to point upwards again in a couple of days 11/12
This visualisation technique is called a connected scatterplot, and it is good for showing the trend of 2 variables over time: it can show delays and cycles really well, as demonstrated http://steveharoz.com/research/connected_scatterplot/">https://steveharoz.com/research/... 12/12 #connectedscatterplot
