One of my pet peeves is seeing a statistical chart used for an inappropriate purpose. Classic examples include bar charts (source) for for bivariate comparisons, line charts (source) for quantitative data by unordered categories, pie charts for anything. So I applaud this attempt to create a simple taxonomy of statistical graphics, to help non-statisticians display data in the most appropriate way.
(Click the image for a larger version, or download the PDF for printing.) Like Gelman, I could quibble with some of the specific choices (waterfall charts?), but overall it makes for a handy reference for choosing a chart according to the kind of message you're trying to convey.
I also like the fact that this chart is just one step (step 7) in a nice process (with a slightly silly name -- wasn't everything "extreme" in 2006?) for constructing and delivering a quantitative presentation. I feel there's often too little consideration of audience, objectives and politics in presentations, so it's nice to see it laid out here. And the advice to not "think in Powerpoint" but to delay writing slides until you've sorted out the other steps is good.
Andrew Abela: Choosing a good chart.
Comments