# Graphs as presentation of numerical information vs. graphs as eye-catching illustrations

I’ve made this point several times recently but just want to get it down one more time.

Kaiser writes that “the gulf between infographics and statistical graphics remains wide,” and he provides some examples of eye-catching charts that don’t do a very good job at all of presenting information.

I think the following two ideas are helpful:

1. Tufte had it write when he wrote of the visual display of quantitative information. Not “data” but, more generally, “quantitative information.” Which can include all sorts of derived quantities, ranging from the Consumer Price Index to logistic regression coefficients (generally best to divide these by 4, of course). Much confusion (not from Kaiser, but from others) has arisen because people think of statistical visualization to be about showing the raw data. No, it’s more general than that.

2. Newspaper and magazine articles are often illustrated by photographs and cartoons which are pretty or shocking or attention-grabbing or interesting and in some way complement the substance of the article. We don’t generally criticize such illustrations as being noninformative; they’re not really expected to convey information in the first place.

I think of infographics (for example this sort of thing featured by Nathan Yau) as a cross between 1 and 2 above. Another example is Harper’s Index (which I always thought was a cool idea, and it even featured my own research once!). It conveys a bit of information but not systematically–it’s more a way to stimulate some thought than to draw any conclusions. It plays the role of a good photographic illustration.

P.S But some displays are cool and informative! I am sad that a cool-but-ultimately-limiting gimmick such as Wordle has become so popular, but I’m happy that there’s also room in the world for something like the Baby Name Wizard, which is cool and also informative and leads to new and interesting ways of thinking about the world.

## 3 thoughts on “Graphs as presentation of numerical information vs. graphs as eye-catching illustrations”

1. I think there are two kinds of "false positives/negatives" here:
– There might be "serious" statisticians who think that only because a graphical display is "cool" it is not conveying real, serious information – this is certainly nonsense, as many examples can show – but of course true for most of the infotainment stuff …
– Even worse is the situation, where statisticians who have little or none expertise in graphing data use "serious" statistical graphics in a completely wrong way and thus can't convey any sensible information either. Unfortunately this is not a too rare scenario.

In the end, the only thing that counts is whether or not a graphical display can make a point. If no cheating is involved, it does not matter how many bells and whistles were used.

Happy Easter!

Comments are closed.