Well, that’s another St. Patrick’s Day behind us. I am not really a green beer person, and don’t care much for corned beef (although I am partial to cabbage). So, I usually don’t participate in the celebrations that occur on the famous saint’s feast day. This year, however, I did make a presentation on St. Patrick and the history of Guinness as part of the monthly Pint Talk series at one of my local breweries, Quenched and Tempered.
As I was preparing my talk, my oldest daughter sent me a text asking if I knew that the student t-test was developed by an employee of Guinness. My daughter is a nurse and is working on her Master of Science degree in Nursing at Ohio University. In one of her classes, the instructor was covering some basic statistical techniques. This is where the student’s t-test came up. I responded to her text, telling her that I was indeed familiar with both the test itself and the fact that it had been developed by a Guinness employee, and that both would be mentioned in my presentation.
William Sealy Gosset was born in Canterbury, England in 1876. As a student at Oxford University, Gosset studied mathematics and chemistry. Soon after graduating from Oxford, in 1899, Gosset joined the Guinness brewery in Dublin, Ireland, as a master brewer and experimental scientist. As it was to turn out, all of Gosset’s working days (1899-1937) were spent as a Guinness employee. When he joined Guinness, it was the largest brewery in the world, producing close to 1.2 million barrels of beer. And it was on a growth trajectory. By 1914, it was producing almost 2.4 million barrels.
The arrival of Gosset at Guinness signaled something of a scientific revolution for the Irish brewery – from that day forward, scientific rigor would underpin all of the brewery’s activities. As noted by Stephen T. Ziliak, “all factors of production, from barley breeding to taste testing” would be “controlled, improved, and confirmed by experimental science”. A degree in science from Oxford or Cambridge would be required of anyone hoping to become a Guinness brewer. By applying scientific principles to brewing, Guinness hoped to improve both the quality and consistency of its beer, while minimizing production costs. One observer likens being a brewer at Guinness in the early 1900s to “being a computer scientist at Bell Labs in the 1970s or an artificial intelligence researcher at Google today“.
Quality and consistency of the final product was highly dependent upon the quality and consistency of the barley and hops used in the brewing process. In the case of hops, Guinness’ traditional method of choosing which to use was based on looks and/or fragrance – a highly unreliable technique. A more reliable method would have been knowledge of the degree of soft resins in the hops. This was problematic, however. The vicissitudes of agriculture meant that there could be considerable variation in the soft resin content of hops, both from field to field and year to year.
Measuring the soft resin content in every hop cone was not feasible. By 1886, Guinness was using ~5 million pounds of hops. An alternative was to take a sample of hops and measure their soft resin content. But this raised some important questions? For example, how many hop cones did you have to sample? And how could you tell if your samples were representative of the larger batch of hops you were considering using in the brewing process. This was one of the tasks to which Gosset set his mind. In 1906 Guinness granted Gosset a sabbatical to work on this problem, so he spent a year at the Biometrics Laboratory of mathematician Karl Pearson at University College London.
Gosset was particularly interested in the problem of small samples. In other words, by measuring the soft resin content of a small sample of hops, can you infer that the entire crop has an acceptable soft resin content? Up until that point, all methods for extrapolating from a sample relied on having at least 30 observations. It was during his year at Pearson’s laboratory that Gosset figured out how characteristics of larger populations could be inferred from a very small number of samples. In his initial work he did not focus on hops. but malt extract. By taking just four samples of malt extract Gosset found that he could be sure with more than 92% certainty that the “degrees saccharine” of the extract was within 0.5 degrees of the required amount which was 133 degrees.
In 1908 Goset published the results of his research in the journal Biometrika. He did so under the pseudonym “Student”. While Guinness was happy for Gosset to share his work, they preferred that competitors not know for whom the author worked and the manner in which the brewery was applying his research. Indeed, Gosset used the same pseudonym in 19 of the 21 scientific papers he published.
As a result of Gosset using the pseudonym in his paper, the statistical technique that he presented became known as Student’s t-test. Today the Student t-test is taught in every introductory statistics course in the world (I don’t know that for sure, but I can’t imagine an introductory statistic’s course without it).
It is also widely used by both academic and non-academic researchers. For example, in medical research it can be used to compare the effectiveness of two drugs (or a placebo and a drug) in a clinical trial, while Psychologists might use it to assess the impact of therapy on an individual (before and after therapy scenarios). A manufacturer may use it to assess whether average customer satisfaction rating for a new product differs significantly from a pre-determined benchmark.
In a letter published in the Pakistan Journal of Medical Sciences, Younis Skaij described the Student’s t-test as the ‘bread and butter” of statistical analysis. Writing in the Scientific American, Jack Murtagh, called the Student’s t-test as “one of the most important statistical techniques in all of science“. As Gosset’s research was published under the pseudonym of Student, he probably does not get the recognition he deserves among the scientific community. But next time you raise a glass of Guinness to your lips, give a thought to this “gentleman scientist“.
Further Reading:
Ziliak, Stephen T. 2008. Guinnessometrics: The economic
foundation of “Student’s” t. Journal of Economic Perspectives, Volume 22, Issue 4, Pages 199–216.
