16 August 2006

Frederick Mosteller: Statistician, Stylometrist, Red Sox fan

The mathematician Frederick Mosteller died last month after a long and rich career. On the faculty at Harvard for forty years, he chaired four different departments and was one of the founders of the school's statistics department. And he put himself through college playing poker!

One of the reasons Mosteller's work is near and dear to my heart is because he was a lifelong Red Sox fan. One of his earlier published works discussed the statistical forces at play in baseball, concluding that the outcome of a game involves so many chance events that the winning team is not always the better team. Hey, Red Sox fans have known that for decades, but Mosteller was the first guy to back it up with science.

I first ran across Mosteller's work reading The Language of Mathematics by Keith Devlin. Devlin describes Mosteller's work on the Federalist Papers - one of the most well-known uses of statistics in linguistic analysis, a field now called stylometry.

The Federalist Papers are a series of essays which appeared in New York newspapers in 1787 and 1788, designed to persuade citizens to ratify the US Constitution. They were authored by James Madison, Alexander Hamilton, and John Jay, although they were published under a single pseudonym. Although both Hamilton and Madison wrote lists detailing who wrote what, the authorship of twelve of the essays is unclear.

Mosteller and his colleague David Wallace used word counts for thirty common words such as "by", "according", "there" and "enough". Analyzing the essays whose authorship was unquestioned, they were able to create a sort of linguistic fingerprint for each author. They compared these profiles to the word counts in the anonymous essays. Although any one word count alone would not yield conclusive results, the thirty words counts together established that Madison was most likely the author of the twelve disputed essays.

Mosteller's research interests were diverse. His career spanned sixty years of active research in a wide range of fields. As a biostatistician, he studied the placebo effect, analyzed the relative risk of the anaesthetic Halothane, and published a statistical risk/benefit analysis of surgery. More recently, he advised the Clinton administration on the impact of smaller classes on children's learning. He called for better use of statistical trials to test innovations in elementary education and in health care, and it's said that his work in sports statistics influenced Billy Beane's innovative management approach with the Oakland A's.

And he finally got to see his team take the World Series in 2004.

No comments: