While you are enjoying a freshly poured pint of Guinness, the last thing you might be thinking about is statistics. Statistics is an important branch of mathematics that have recently re-incarnated into fields like data science and artificial intelligence. This is a story about how a brewer in the Guinness factory in Dublin, Ireland inventing a widely used statistical method while trying to select the best yielding varieties of barley for brewing. It is also a story of how Guinness did not allow him to publish his findings under his name, and how the innovative brewer got them published under the pseudonym of “Student.”
Student’s t-tests are a widely used method in statistics to compare two or more data-sets, in relation to the variation in the data. There are three variants of Student’s t-test, the one-sample t-test, the unpaired t-tests and paired t-test depending on the nature of the data-sets. These methods are important because when statisticians and data scientists first come across a large data-sets they typically qualitatively compare them to see how co-related the data-sets are. Most statistical packages, Python, R, Microsoft Excel, SAS, SPSS, Stata, Matlab and Minitab, include implementations of Student’s t-test.
This statistical method was discovered by William Sealy Gosset who was born in Canterbury, England in 1867. He studied mathematics and chemistry at New College, Oxford, and then started working as a brewer for Guinness Breweries in Dublin in 1899. In 1900 Guinness opened the “Guinness Research Laboratory,” under the leadership of Horace Brown. Brown was a renowned chemist, who was interested in using scientific methods to standardize the quality and costs of many varieties of barley and hop used to brew Guinness.
In 1907 Gosset was given charge of the Guinness’ Experimental Brewery, where he used his experimentally derived “Student table” to determine the best variety of barley to be used for brewing. He collaborated with Karl Pearson between 1906-1907 and developed his theory of estimating errors in small sample sizes. Karl Pearson founded the world’s first university statistics department at University College London in 1911 and contributed significantly to the field of biostatistics. In his pioneering paper in 1908 Gosset addressed the brewer’s concern with small samples, biostatisticians like Pearson, on the other hand, typically worked with large data sets and were not interested in developing small-sample methods.
However, Guinness had a policy that prohibited its employees from publishing any academic papers. This was a direct result of another industry researcher while working in the Guinness Research Laboratory publishing a paper containing Guinness’s trade secrets. However, Gosset was persistent and kept pleading with the upper management explaining how his statistical and philosophical conclusions were of little use to competing businesses. The upper management finally allowed him to publish under a pseudonym “Student” to prevent discourse among the rest of the research staff. As a result, Gosset’s methods are commonly known as Student’s t-test.
Gosset left Dublin in 1935, to take up the position of Head Brewer, at a new Guinness brewery at Park Royal, London. Gosset died two years later in 1937, and the rest of his work was also published under his pen name of “Student” (‘Student’s’ Collected Papers (edited by E.S. Pearson and John Wishart, with a foreword by Launce McMullen), London: Biometrika Office. (1942)). Gosset’s work was not at the forefront for many years and was primarily used by researchers at Guinness’s breweries. However, Gosset had sent his work to R. A. Fisher, who was at that time working at the Rothamsted Agricultural Experiment Station. It was Fisher’s work on generalizing Gosset’s theory, that later helped the Student’s t-test to gain its popularity among statisticians.
Industrial researchers like Gosset with a strong urge to solve a real-world problem will continue to deliver the most impactful inventions. However, still today, one of the biggest dilemmas for industrial researchers is whether to publish their know how or to keep it a trade secret for the sole benefit of their corporation. Gosset’s strategic collaboration with university researchers popularized his work. Perhaps the most important take-home message from this story is that today’s industrial researchers must have a clear strategic roadmap to share their knowledge with their peers in the academia and industry, for science to progress in its due course.