Any statistics treatise that includes a reference to "the great experimental maltster and barley farmer" Edwin S. Beaven is obviously well worth reading. Naturally, I had never heard of Mr. Beaven before imbibing "Great Lease, Arthur Guinness -- Lovely Day for a Gosset!", Stephen T. Ziliak's tale of how a master brewer at Arthur Guinness Son & Company, Ltd, perfected "small sample theory." I'm just happy to know that there is a world where experimental maltsters get the recognition and hero worship that they deserve. This is a good thing. (Found via Mark Thoma.)
There is much to learn from Ziliak's mildly inebriated story of "Guinessometrics," to be published in the Journal of Wine Economics' special "Beeronomics" issue. For example, I did not know that Nigeria is Guiness's third largest market after the UK and Ireland. And I was sadly unaware that "Fermentation studies have been at the center of quantitative progress in science."
Although it sure makes sense when you think about it. If necessity is the mother of invention, then the necessity of brewing quality spirits should naturally be the mother of quantitative progress in science. Of course Joseph Priestley discovered carbon dioxide while "studying gases emitted from experimental beer he was brewing." And of course Lavoisier's brilliant insight into "the principle of conservation of mass" came in the context of his description of what happened when he experimented with fermenting grapes into wine. Who says science and alcohol don't mix? Without alcohol, we might not have science. Or at the very least, chemistry.
I am not well enough versed in the history of statistics to judge Ziliak's central thesis, which is that the legendary master brewer William Gosset's "exacting approach to discovering the sources of errors, both random and real, in his and in others' experiments on fermentation and plant inputs, marked a significant advance over the approaches taken by previous students." But packed in with so many jokes about hops and yeast, one hardly wishes to quibble..
Say goodbye to the "look-touch-and-sniff approach of 18th century craft brewers." By the time Gosset arrived, in the early 20th century, Guinness was the largest brewer in the world, and Gosset had to solve some hard practical problems. For example, when producing beer by the millions of hogsheads, how do you ensure the quality of your ingredients when dealing with huge quantities? As Ziliak notes: "You don't have to tell a brewer that running a series of experiments on malt characteristics or of yield trials on barley can grow expensive, fast." You can't test it all: In fact, there's an economic imperative mandating testing as small a sample as you can get away with. But the smaller the sample tested, the larger the potential error rate. According to Ziliak, Gosset devoted four decades to refining the state-of-the-art in statistical analysis on such subjects, all for the greater glory of Guinness.
But then again, how does such brilliance compare with a can of Coors Light featuring mountains that turn blue when the "world's most refreshing beer" is appropriately cold? How come Gosset never thought of that?