Leland Wilkinson points out that flexible, generic, plotting programs, such as SigmaPlot, allow one to make “ungrammatical” plots. That is, one could take a histogram of a log of one’s data; and then find the mean. That is meaningless, he points out. He persuavely argues that one can define a strict “grammar” of procedures to avoid such “meaningless” visualizations due to unthinkingly swapping Two summarizing quotes stand out:
I assert the Grammar of Graphics is THE grammar of statistical visualization.
There is always a model. In any statistical visualization, we always have a model.
What is it? What is the mathematical model?