Edward Tufte pointed me to this article, looking back on his papers from the 1970s on midterm elections and seats-votes curves.

I’ve been interested in seats and votes since I was in grad school in the 1980s and so I was aware of his poli sci papers at that time; coincidentally, my brother gave me a copy of his first book back in the 80s because he just thought I’d like it. My favorite Tufte book is his second, because I’ve been very strongly influenced by the “small multiples” idea (which I’ve since been told has appeared in the work of Bertin as well, and then Tufte gave a reference to Scheiner from 1613). I used to love graphs with many lines on a single plot (or many different points, using symbols and colors to distinguish), but now I like to use small multiples whenever possible, even using somewhat arbitrary divisions to make graphs clear; for example, see Figure 2 here or, for a purer example, Figure 1 here.

My recent research involves using graphs to understand fitted models. Statisticians tend to focus on graphs as methods of displaying raw data, but I think models themselves can be better understood graphically; for example, in our red-state, blue-state paper.

My only Tufte-related story is from our paper on reversals of death sentences, Figures 2 and 3 of which show the “pipeline” from homicides, to arrests, to prosecutions, to sentencing, to judicial review, to execution. My collaborator had a friend, a graphic artist, who wanted to make a “Napoleon’s march”-style graph of the death sentencing pipeline. It didn’t look so good, however. Two problems: (1) no geographic dimension, which is what made the Napoleon graph so distinctive; (2) lots and lots of missing data (cases still under review) which throws off the picture. So in the end I was happier with the flowchart-like diagram we made. Although I’m sure something better could have been done.