I couldn’t wait to get to the Guinness brewery after landing in Dublin. Yes, I was eager to taste a pint nearly fresh off the line, but I was also curious to see if there would be any indication that the brewery was home to arguably one of the most important developments in the field of statistics. I was not disappointed.
There on the wall in the old Guinness Storehouse was a lone plaque paying homage to brewer, chemist, and statistician, William Sealy Gosset. Unfortunately, none of the other tourists had a clue why my husband and I hurried toward it to snap some photos. But I digress....
More than a century ago, and well before the dawn of Six Sigma, Guinness understood the importance of quality control. Starting in the early 1900s, Guinness began recruiting scientists as master brewers to apply scientific methods to beer production. To brew the perfect pint, Guinness needed to make sure that the ingredients used to brew the beer met the company’s high standards. These ingredients included water straight from the Wicklow Mountains, hops, yeast, and of course, the barley that gives Guinness its rich distinctive color and taste.
The statistician William Sealy Gosset was tasked with studying and testing these essential ingredients to ensure that every pint was brewed to perfection. However, back in his day, statistical tests for averages were traditionally conducted using large sample sizes rather than the small, cost-effective samples that Gosset was using. And Gosset recognized that statistical methods based on large samples would not work for small-sized samples.
What’s a statistician to do? Gosset went to work and devised a statistical method to effectively analyze these small samples and their means—and thus the t-test was born.
Guinness understandably didn’t want competitors to know it was using statistics to make a better brew, so Gosset’s findings were published under the pen name “Student.” And so we have what is known as the Student t-test and not the Gosset t-test. Who knew the statistical world could be associated with a little bit of mystery and drama—or beer for that matter?
Now that you know the history of the t-test, I’ll let you in on a little secret: Guinness really does taste better in Ireland. I can attest to this from personal experience, the Guinness overall enjoyment score (GOES), and the t-test support my claim.
Using the GOES rating system (aka the visual analogue scale), in 2011 four researchers—one each from Ireland, England, Holland, and Germany—traveled to pubs around the world collecting data on the taste of Guinness. What a rough job that must have been! Anyway, the researchers noted that pints consumed in Ireland received a mean GOES score of 74, while the average GOES score for Guinness tasted elsewhere was 57. Their findings appeared in the March 2011 issue of Food Science.
Is this difference of 74 vs. 57 significant, or is it simply due to natural, random variation? Bring out the t-test. Similar to what the four researchers must have done, we can use Minitab software to run a t-test to analyze these averages. Suppose this is a summary of the data that were collected.
With a p-value less than 0.001, the researchers were able to reject the null hypothesis (i.e., no signiicant difference in the specified populations) and conclude that, in fact, Guinness served in pubs in Ireland tastes significantly better than the pints served elsewhere. Brilliant!