Yesterday I bottled a Guinness-style dry Irish stout I named “Student’s Guinness-Style Irish Stout.” I’m sure fellow statistics geeks out there realize why but I thought I’d pass on the reason for the name.
In the late 1800’s a fellow stats geek named William Gossett was hired to work in Guinness’s pilot brewery evaluating ingredients. Most of the statistical methods we use today hadn’t been invented so he had no easy answer to how do you determine difference between ingredients given you can do a very limited number of pilot brews with them. As today, you couldn’t tell a lot about ingredients from their statistics, you had to brew with them. This meant his sample sizes were very small, five or less. To use the normally accepted z-test required 30 samples. So Gossett developed the math to detect difference in small sample sizes.
Then, the knowledge Gossett had developed was a closely guarded trade secret – Guinness was the only brewery that knew how to do the test. But Gossett, a scientist, also had the right to publish so to guard his true identity and that of his employer, he employed the pseudonym “Student.”
The test, the bane of many elementary statistic students, is called “Student’s t-test.” It’s easily the most famous statistical inference test and it has beer in its pedigree. Cheers!
Forgot to mention brew day yesterday, Belgian Saison. More to come later.