1. In tracking #COVID19 I make frequent use of @jburnmurdoch's semi-log plots of daily death by country. A semi-log plot is helpful because a slope on a semi-log graph corresponds to an exponential growth rate. By looking at changes in slope, we can see changes in R0 at a glance. https://twitter.com/jburnmurdoch/status/1247666977566490626

2. But be cautious when you try to interpret these graphs.

Before you read on, take a look at the example below. How you would interpret the trajectory of these two countries? Is country A flattening the curve? How about country B?

3. It sure looks like they both are flattening the curve, and one feels optimistic looking at both trajectories.

But this not correct, because concavity on a semi-plot does not imply concavity on a linear plot.

(Don't worry if that doesn't make sense. Just read on.)

4. Here are the same two countries with the same two trajectories, plotting on a linear scale. County A flattened things out a bit but deaths are still rising.

I wouldn't say that country B is flattening at all—deaths are rising, and they're rising more and more each day!

5. Wild, right?

That's why you have to be really careful in how you interpret data graphics. I love the @jburnmurdoch graph that started this thread. Nothing wrong with it. It's great dataviz. But a single visualization can't do everything.