The Pleasures of Statistics is the autobiography of Frederick Mosteller (1916–2006). Mosteller stopped work on his manuscript around the time of his retirement in 1990 and gave it to editors Stephen Fienberg, David Hoaglith, and Judith Tanur in 2003.
The autobiography is divided into three parts. Part I is devoted to six studies for which Mosteller is best known:
- An analysis of the pre-election polls of 1948
- The Kinsey Report Sexual Behavior in the Human Male
- Foundations of mathematical psychology
- Authorship of the disputed Federalist Papers
- The safety review of the anesthetic halothane
- The Coleman Report on educational equality
These chapters contain few statistical details. Instead, they give the story behind the studies and explain in simple terms some of the difficulties involved. Here we will look at just the study of the Federalist Papers, a collection of 85 articles written by Alexander Hamilton, James Madison, and John Jay. The Federalist Papers were published anonymously, though the authorship of most of the articles became known after Hamilton's death. The authorship of twelve of the papers was still in dispute when Mosteller became interested in the problem in 1959.
Scholars had used statistical methods to resolve previous authorship disputes. One technique was to look at sentence length; some authors tend to writer longer sentences than others. This approach was useless for studying the Federalist papers because Hamilton and Madison used identical sentence lengths, 35.55 and 35.59 words respectively. Variation in sentence lengths was also no help. The standard deviations of sentence length were 19 and 20 respectively for the two authors.
The initial investigation of the Federalist Papers was entirely manual. The articles were typed and cut into individual words so they could be sorted and counted. The analysis of the word frequencies, however, was carried out with computers. This was one of the first major studies to take advantage of computers.
The statistical analysis of the Federalist Papers was also one of the first major studies to use Bayesian statistical methods. In fact, Mosteller became interested in the authorship question because he was looking for an application of Bayesian statistics. The study pioneered methods that were influential in authorship analysis as well as in statistics more generally.
Part II of The Pleasures of Statistics is primarily biographical. This section is brief and focuses primarily on Mosteller's education. The author emphasizes the elements that were relevant to his eventual career as a statistician. His frustrations with foreign languages along the way are amusing to read.
Mosteller describes a sort of mathematical epiphany when a mentor, E. G. Olds, showed him generating functions, a method Mosteller calls "magical."
Although I had loved mathematics all along, this was the first time I ever felt that I'd been working with a peashooter when I could have had a cannon, and furthermore, that the tricks I had learned could produce such a marvel.
He mentions that he enjoyed playing poker; perhaps this lead to his interest in probability and statistics.
Part III of The Pleasures of Statistics is a series of topical chapters rather than a chronology. One gets the impression Mosteller was running out of interest in writing a more detailed autobiography. Here he gives his thoughts on topics such as writing, collaboration, and health policy.
Frederick Mosteller had long and productive life. His autobiography gives a glimpse into his work in teaching, statistical research, and public service. He remained active in his retirement years. The achievements of these years are discussed in an epilogue by the editors.
John D. Cook is a research statistician at M. D. Anderson Cancer Center and blogs daily at The Endeavour.