We asked Andrew Gelman (@StatModeling), a professor of statistics and political science at Columbia University, to tell us about his favorite scientific plot. He chose a pair of figures from his 2010 article Public opinion on health care reform, published in The Forum.
Each figure reports the result of a survey, administered once in 2000 and again in 2004, that asked people whether the federal government should spend more money on health care for the uninsured. The 4 × 5 grid of maps of the contiguous United States reveals the spatial distribution of answers, broken down by household income (column) and age (row). By using the technique of small multiples, popularized by Edward Tufte, the plots each display 48 × 4 × 5 = 960 numbers, yet still appear clean and organized. Gelman likes the plots "because they allow comparisons in many dimensions at once."
The plots are also notable for using a sensible color scheme. Neutral grey is the average answer across the entire sample; darker teal is more agreement (yes on spending) and darker tan is less (no); white means that there was little data.