stats n' beer

if you ever interview at minnesota, you can expect to spend a lunch hour discussing teaching with faculty and grad students. i was initially skeptical about this practice, but now i really look forward to the discussions. it is typically a loose, free-form hour, but it helps us communicate that we actually like teaching here and that we especially like rock star researchers who are thoughtful about teaching.

this week, we got to discussing social statistics. i always like to sprinkle my crim lectures with tales of sutherland or hirschi or the convention bar tabs of certain criminologists. i know fewer stories about statisticians, but when i taught intro stats i loved the one about the inordinately useful student's t-distribution.

a real statistician can correct me on this, but i believe that the t-distribution was introduced in biometrika in 1908 by william gosset, who worked as a brewer for guinness in dublin.* he adopted the pseudonym "student" because the brewery was concerned about revealing trade secrets in print. mr. gosset's work was popularized in part by the great r.a. fisher, who referred to the t-distribution as student's distribution. the story is better told elsewhere and there are a few nice wiki versions floating around as well.

i have it on good authority that some soc and crim students enjoy a frosty beverage from time to time, so they might appreciate such nuggets. i'd like to know precisely how the brewery plied these statistical tools, however, to flesh out the story. comparing the purity levels of beer from two separate vats? i'm also intrigued that claude guinness would recruit the best and brightest from cambridge and oxford to do brewery work, but perhaps this practice helps account for guinness' great success over the past century.

*Student. 1908. "The Probable Error of a Mean." Biometrika 6: 1-25.