John Tukey wrote about semigraphic displays. I think his most famous effort in that area–the stem-and-leaf plot–is just horrible. But the general idea of viewing tables as graphs is good, and it’s been a success at least since the early 1900s, when Ramanujan famously intuited the behavior of the partition number by seeing a table of numbers and implicitly reading it as a graph on the logarithmic scale.

To return to the present, Steve Roth sent me a link to these table/graphs that he made:

Europe vs. US: Who’s Winning? and State Taxes and Prosperity, Revisited. He writes:

I [Roth] find the layout with the red/black gives a simultaneous numeric and graphical representation of the situation, and condenses a lot of immediately apprehensible info into a small space.

It also helps me avoid at least one axis of cherry-picking (periods), which I am as prone to as all humans are.

Any thoughts welcome. In particular, do you think the average and count aggregates at the bottom of the second post are of any value?

I’m also wondering if your travels through econometrics have yielded the same sample-period-related frustration that I’ve felt with most of the research.

I do like these displays, which are designed for the sort of question I haven’t ever thought much about, which is how to be fair in displaying comparisons over time.