October 7, 2015 at 10:42 AM by Dr. Drang
Occasionally, I write a post about making charts. Sometimes these posts are rants about poor practice or my thoughts on good practice, but usually they’re more descriptive than prescriptive. I write about how I make charts with the expectation that my beliefs and tastes will come through and that I might have some small influence in stemming the tide of bad charts.
A layout that might work with a few candidates is a mess with fifteen. The legend overwhelms the chart, and there’s no rhyme or reason to the order of the names. The colors are too close to one another. The markers, which could be used to distinguish the candidates, are the same for each.1 The labels for the horizontal axis are stupidly formatted over two lines. Worst of all, the polls are equally spaced horizontally even though the times between them vary from 5 to 14 days.
You might say this is nitpicking and that the important thing is that the chart communicates who’s winning and who’s moving up or down in the polls. You could also argue that there’s no reason to wrote an article with correct tenses or to gets the verbs to agree with the subjects. Them things isn’t important to communication, is they?2
Into this mess steps Kieran Healy, associate professor of sociology at a basketball university down in North Carolina (no, the other one). Kieran is perhaps best known on the internet for a data visualization post that has, unfortunately, become something of an evergreen. His charts are always tasteful and informative because he’s a smart guy and he’s thought a lot about how to communicate through plots.
This semester—half-semester, actually—Kieran’s going to impart his wisdom to grad students in his department through a special topics course. He’s starting out right, by demonstrating the evils of Excel’s overly cute 3D column charts:
A handful of Duke sociology students won’t fix the world’s data visualization problems, but Kieran is making his class notes available on GitHub, so there’s hope that others will find them and learn.
Distinguishing the data series is somewhat easier in the actual chart (as opposed to this screenshot) because you can click or tap the names in the legend and see the corresponding series light up. Of course, you have to know or guess that this is possible, otherwise you’d never try it. ↩︎
Given my penchant for leaving typos and editing artifacts in my posts, this is a very dangerous paragraph. ↩︎