Junk Charts is a great blog that takes an example of a data visualization, critiques it systematically, and then either improves it or shows a different way of displaying the same data. The site doesn’t go for overly elaborate graphics, just clear and effective ones. This post has a roundup of the most viewed posts and the author’s favorite posts of 2020.

One thing you probably shouldn’t do is describe interesting graphics in words. Nonetheless, here is some data, which I am not putting in a visual form because it would take exponentially longer than just listing it out:

• There are 12 graphics covered by the post.
  • 2 scatter plots
  • 3 bar charts
    • 2 horizontal, not stacked – one of these gets changed to a bump chart
    • 1 horizontal, stacked – actually this is more of a “tree plot” where two data points are stacked and then a third is placed underneath
  • 2 pie charts
    • 1 3D pie chart – gets converted to a bump chart
    • 1 is allowed to continue to exist as a pie chart, with minor tweaks
  • 1 “dot matrix” (I’m not even sure if this is the best name, but basically you have empty squares or circles showing the total number of a thing, then some of them get filled in to illustrate how many of that thing fit a certain category)
  • 3 time series plots
    • 2 conventional – although one has two vertical axes, and the author illustrates how the limits can manipulated to suggest to the eye that two trends are related, or not
    • 1 showing shaded regions over time – basically a stacked bar changing over time
  • 538’s election snake

There is something intuitive about pie charts – that is why we explain fractions and percentages to children in terms of pizza or pie, and they grasp it instantly. Pie charts are obviously the wrong way to compare the absolute magnitudes of things.

I do like tree plots. I made one in 2020 and I was proud of myself – it showed the number of acres served by stormwater management controls implemented by three different administrative programs. And then I made a second one where I broke the numbers down further within each of the categories. This was very effective in conveying how much is actually achieved by each of the programs compared to the effort and expense that goes into them.

Resolution for 2021 is to play with “dot matrix” plots at some point (and maybe learn what the best name for these is.) I think these are effective in putting numbers in context of bigger numbers, regardless of units. For example, my city has around 80,000 cumulative confirmed coronavirus cases, maybe 5,000 confirmed active infections (about the number of confirmed cases in the last 10 days), maybe between 80,000 and 800,000 actual cumulative infections, and a population of about 1.6 million. I don’t know how many have been vaccinated at this point, but probably a few thousand. So maybe I would make 16 or 160 boxes each representing a chunk of people, and start coloring them in. Then we could see at a glance how much of the population might have some immunity to the virus right now, and how much does not. You could slice and dice the data many ways. Of course, some people died or moved away, and others were born or moved in. Incidentally, about 2,600 people died of Covid, 400 were murdered, and 120 died in and around motor vehicles. I haven’t seen numbers on suicides or drug overdoses but they are always horrifying. Around 1% of any given population dies in any given year from a combination of preventable and not preventable causes, which is sad but news flash: we are mortal beings.

This site doesn’t do maps, which is fine. I am a big fan of maps. But I have a very simple test – is the data geographic in nature? Then make a map. But often, some other types of graphs and tables will further illuminate the data, and those often work well alongside your map rather than being shoehorned into your map where they don’t really belong. And I also find it clunky trying to do any type of mathematical analysis in mapping software when the analysis is not spatial in nature.