This is the chart I've wanted to see: growth in COVID cases in a metro by prevalence, not time. If you'll indulge me, let me explain why
nytimes.com/interactive/20…
nytimes.com/interactive/20…
1) Why metropolitan areas, not states/countries? It's the natural extent of both rapid community spread and health resources.
Saying NYC > Italy per capita isn't a great comparison; huge parts of Italy unaffected, NYC still better than Lombardy, which is what 'Italy' evokes
Saying NYC > Italy per capita isn't a great comparison; huge parts of Italy unaffected, NYC still better than Lombardy, which is what 'Italy' evokes
2) Why a chart of *growth rates*?
Everyone's interested in 'bending the curve.' You can guess it by eyeballing it on a chat of cases. But a chart of rate of growth = just a direct measure of whether the curve is flat or steep
Everyone's interested in 'bending the curve.' You can guess it by eyeballing it on a chat of cases. But a chart of rate of growth = just a direct measure of whether the curve is flat or steep
3) Why growth rate at a certain prevalence, rather than time?
This might take more than one tweet.
Time is the natural one, I get it. But with growth, the question is not *when* rapid growth happens, but whether growth is happening on a base of large N
This might take more than one tweet.
Time is the natural one, I get it. But with growth, the question is not *when* rapid growth happens, but whether growth is happening on a base of large N
Think about the chart that everyone shows, of cases over time, starting at x cases.
On that chart, your eyes are drawn to the country that's highest. You might assume they're worst. But that's not true, if they've since slowed the growth in cases.
On that chart, your eyes are drawn to the country that's highest. You might assume they're worst. But that's not true, if they've since slowed the growth in cases.
Similarly, you might assume that a country that appears 'lower' than other countries on that chart is doing fine. That's not necessarily true, either, if they're experiencing rapid growth after initial slow growth.
This chart--growth over prevalence--fixes this common misinterpretation.
On this chart, the metro that's highest at a given value of x is the one that's doing the worst job of "flattening the curve" at a comparable stage of its outbreak. So it's visually intuitive
On this chart, the metro that's highest at a given value of x is the one that's doing the worst job of "flattening the curve" at a comparable stage of its outbreak. So it's visually intuitive
Anyway, I don't think this chart obviates the need for the simple chart of cases over time. And it is complicated, I get it. But I do think it directly offers much of the information that people are trying to glean for themselves out of cases over time.
That doesn't make it perfect. I wish we had growth in active cases, not cumulative cases. But we don't have that data in a comparable way across geographies. And at this early stage of the crisis, most cases remain active so it's not a big deal (though not hold over long run)
I also wish we weren't limited by varying testing across geographies. Obviously that's a problem with everything right now. I don't agree that this irrelevant or not worth anyone's time. We often have to deal with imperfect information, but that doesn't make it useless
