Read on Twitter
#Coronavirus deaths are often presented in this form, as total deaths vs time, using a logarithmic y axis. The x coordinates are shifted to allow comparison between countries at different stages of outbreak. It is fine, but it is not ideal. (1/9) https://twitter.com/faisalislam/status/1247562441581629442
Here the x axis shift is done by the "from 50 day measure", which is an arbitrary choice and means that the curve positions depend strongly on what happened early on in the outbreak, weeks ago, even though that is not really relevant anymore. (2/9)
Of course that is just a reflection of the fact that in this plot you should not be comparing curve positions but their gradients. However, I think most people instinctively compare the curve positions anyway, and actually the gradients are difficult to read from the plot. (3/9)
Therefore, it would be better to plot the gradients instead, i.e., the growth rate. And to avoid the need of an arbitrary x axis shift, plot it against the total deaths. Like this: (4/9)
The y coordinate shows the rate of growth of the death toll. If the growth was exponential, these curves would be horizontal. A curve turning down means the growth is slowing down. When it hits the x axis, as for China, there are no more new deaths. (5/9)
This shows that in Italy and Spain the outbreak has slowed down significantly. In the UK, USA and Germany there are early signs of slowing down. (6/9)
In fact, to compare the impact of #coronavirus on a country, it makes sense to use deaths per capita, or deaths per million people, instead of absolute numbers. Like this: (7/9)
Either way, the plots shows that the UK is in a worse position than Italy was at the same stage of the outbreak, but in a better position than Spain was. The data is the same as in Whitty's plot, but plotted this way it is a lot easier to see. (8/9)
In fact, this is how we cosmologists look at the early inflationary expansion of the Universe, which was also nearly exponential. More specifically, these plots correspond to plotting the Hubble rate vs e-foldings, which is a lot more informative than scale factor vs time. (9/9)
