Why We Might Not Have Statistics Without Guinness Brewery – A History of the t-Test (10-0)

Why We Might Not Have Statistics Without Guinness Brewery – A History of the t-Test (10-0)


As students of statistics we owe a great debt
or gratitude to our friends at the Guinness Brewing Company. You see there’s a lot of
science that goes into brewing beer and in 1903 Guinness got all sciencey in creating
in creating an experimental malt house at their brewery at St. James Gate in Dublin,
Ireland. At this experimental malt house the brewers slash scientists could grow all varieties
of barley from seed corn to harvest, to malting, to brewing, to the final glorious product.
In nature barley has a lot of variability and one of these scientist slash brewers named
William Sealy Gosset was very concerned about the variability of the barley plants because
you can’t brew a consistent porter or stout if your barley is different from one season
to the next. At the time, statisticians like Karl Pearson and Sir Francis Galton were using
huge sample sizes which meant that the parameters they were estimating approached a normal curve
but for a young brewer estimating barley yields the sample sizes were small maybe three or
four. And that is a problem because small sample sizes did not look like a normal distribution
but fortunately William Sealy Gosset solved this problem for us. He created tables now
called T-tables that account for the variability of small sample sizes. Of course, he wanted
to publish his tables but then — so here’s where the stories diverge. Some people will
tell you that Guinness would not allow any of their employees to publish for fear of
divulging trade secrets. Other people will tell you that Gosset publish anonymously and
that Guinness never found out until after his death but according to the minutes of
the Guinness board Gosset was granted permission to publish with the stipulation that he use
a pseudonym and so in 1908 William Sealy Gosset published two articles in Karl Pearson’s journal
“Biometrica” under the pseudonym Student and today we still use Student’s T-table whenever
we do a T-test. So let us lift a glass to William Sealy Gosset who proved that a small
sample size is nothing to be ashamed of and some great things can come from statistics.

Leave a Reply

Your email address will not be published. Required fields are marked *