Gas station without pumps

2020 April 1

Best visualizations of the COVID-19 spread

Filed under: Uncategorized — gasstationwithoutpumps @ 10:03
Tags: , ,

This is a followup on my post Exponential and logistic growth, which was intended mainly as as teaching opportunity for showing the value of log scales on graphs.

In the comments, Miguel Aznar pointed to videos from MinutePhysics (https://youtu.be/54XLXg4fYsc) and TomRocksMaths (https://youtu.be/NKMHhm2Zbkw), and whatisron pointed to https://aatishb.com/covidtrends/ and https://www.youtube.com/watch?v=gxAaO2rsdIs.

The best static visualization I’ve seen is at http://91-divoc.com/pages/covid-visualization/, which allows you to choose log or linear y-axis scales, plotting confirmed case, active cases, new cases/day, deaths, or recoveries, and (most importantly) having a choice of plotting either raw numbers or normalized by population size.  The normalization by population size is important for comparing efficacy of different approaches, as the total numbers mainly tell you how big the country or state is, and now how much of an impact the COVID-19 pandemic is having.  All the plots have time on the x axis, but start the clock at different times for different countries or states, with time=0 being where the case count=20, death count=5, case rate=1/million, or death rate=1/million.  One could probably get a denser clustering of the curves by being more sophisticated about the definition of time=0, but this method has the advantage of simplicity.  Another nice feature of this visualization is that you can choose which country or state to highlight, so you can, for example highlight your own state to see how it compares with the cloud of others.

Some very interesting outliers are the countries with very slow spread (Taiwan and Japan), or initial rapid spread followed by shutting down the spread (China and South Korea).  Iraq has har fairly slow spread, but Spain and Turkey have had extremely rapid spread.  The United States has pretty much been following Italy’s curve for confirmed cases, but fortunately not for deaths (Italy is at over 200 deaths per million and has not plateaued yet).  Spain and Belgium have had the fastest growth in per-capita deaths, and Spain may still overtake Italy—both are at the point where their health-care capacities are exceeded.  Taiwan and Japan have been so successful at slowing the spread of cases that they don’t even appear on the death plots, not having reached the 1 death/million threshold that the graph makers were using.

If the US reaches 200 deaths/million (as seems likely given how we are following Italy’s curve for per-capita cases, we can expect to see 66,000 deaths in the US.  Since Italy’s death toll is still going up fairly rapidly, and there is currently no evidence that the US is doing any better than Italy at slowing the spread of COVID-19, we can probably expect 2–3 times that death toll (consistent with the optimistic scenarios from the US government).

At the state level, New York, Michigan, and New Jersey  stand out for very rapid growth of cases and of deaths (so it isn’t just more testing).  Oregon and Vermont stand out for slow growth of deaths.  California is low for confirmed cases per capita, but middle of the pack for deaths per capita, so California is probably way behind on testing.

5 Comments »

  1. Hi Kevin

    In looking at http://91-divoc.com/pages/covid-visualization/ in the chart COVID-19 Cases by Country, the US is shown between “doubling every two days” and “doubling every three days”, but on looking at the data, there were 101,608 cases on day 23, and doubling happened between day 27 (188,123 cases) and day 28 (213,323 cases) — this looks like doubling took about 4.5 days. This is consistent with the daily growth rates given.

    The good news about the chart is the ability to switch between log and linear scales (the latter makes it very easy to see there is something wrong).

    Or perhaps “misleading”. The “doubling every 3 days” dotted line is more acute than the US data so it is not “wrong” but what is needed for clarity would be to extend the US with a dotted line labeled “currently doubling every 4.5 days” or some such.

    I think it would be even more clarifying to have a number of labels along a country line that tag the doubling rate every 4 or 5 days …

    Cheers

    Alan

    Comment by alanone1 — 2020 April 1 @ 23:05 | Reply

    • I think your confusion comes because the dotted lines are only to show the slope of different growth rates. The slope for the US is currently doubling about every 4.5 days, but it had a much higher slope earlier, so the long-term average growth rate is faster than doubling every 4.5 days.

      There are tags on every data point that give the growth rate, though in growth per day rather than doubling time.

      Comment by gasstationwithoutpumps — 2020 April 2 @ 07:54 | Reply

      • I wasn’t confused. I could see what they were trying to do, and of course I examined the datapoints.

        The problem is that this is poor graphical/UI design — especially given that the graphs are active and running in the browser on top of JS.

        If Ed Tufte is the patron saint of good informational graphics design, then a “mortal sin” would be mislabeling the axes, and a “venal sin” would be to do what was done here.

        Comment by alanone1 — 2020 April 2 @ 08:13 | Reply

        • I’ve read Tufte’s books and gone to one of his revival meetings. He’s got a lot of good advice, though I don’t agree with his taste on everything.

          I agree with you that the long dotted lines to show slope can be misleading—but I’ve not thought of a better way to show so much data without it being even more confusing. Projecting the current growth rate into the future is likely to be even worse—particularly for some of the noisier curves. Maybe it would help to draw the slope lines centered at the most recent day of the highlighted curve, rather than day 0 of all curves, or to move them around with the cursor.

          Comment by gasstationwithoutpumps — 2020 April 2 @ 08:45 | Reply

          • Yes, I think your suggestions would help. Going from doubling every 2 days to every 4.5 days in a few weeks makes showing an average of the whole progression to be a not very useful measure. I think something that shows the progression of doubling over time would be very interesting (and maybe using distinct colors so the superposition is not confusing).

            Comment by alanone1 — 2020 April 2 @ 09:55 | Reply


RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: