Mathematical Theorist Frederick Mosteller
Tuesday, July 25, 2006
Frederick Mosteller, 89, who founded Harvard University's statistics department and used mathematical theories to explain everyday concerns, from health care to the World Series, died July 23 at Powhatan Nursing Home in Falls Church. He had sepsis.
Dr. Mosteller was a premier statistician of his generation and an early promoter of methodologies that can affect public policy, including analysis of how students can learn better and how some surgical practices can improve lives. He worked across many disciplines, wrote hundreds of papers and shared dozens of book credits with authors such as Princeton University statistician John W. Tukey and future U.S. senator Daniel P. Moynihan (D-N.Y.).
Dr. Mosteller worked with Moynihan, then a Harvard government professor, on comprehensive studies involving the effects of a child's home life on educational development. Their book, "On Equality of Educational Opportunity" (1972), argued that raising the income of families was far more effective in elevating academic achievement than plowing more money into schools.
His work in meta-analysis, a comprehensive study of other studies, was also notable. In 1995, he strongly supported a Tennessee educational study showing that smaller class size vastly improves the rate of learning for young students. His imprimatur reportedly helped spur President Bill Clinton's request for 100,000 new teachers to reduce classroom sizes.
In 1962, Dr. Mosteller also found himself in the news when one of his studies addressed the foundations of U.S. history.
Dr. Mosteller and a colleague from the University of Chicago, David Wallace, proposed a solution to a lingering mystery of political science: Who had written 12 of the 85 Federalist papers?
Those essays appeared in New York newspapers in 1787 and 1788, most under the pen name Publius, to urge the ratification of the U.S. Constitution. Although James Madison, Alexander Hamilton and John Jay were known to be "Publius," it was unclear which of the three had written a dozen of the pieces.
Madison and Hamilton were assumed by many to be the authors; Jay, who became the first chief justice of the United States, had never been a likely candidate.
Dr. Mosteller and Wallace spent three years on the project, applying Bayes' theorem, a method of interpreting the probability of one event based on previous experience of other connected events. They also had at their disposal a high-speed IBM computer at the Massachusetts Institute of Technology, which they fed the known Federalist writings of Madison and Hamilton.
Among other things, they looked at sentence length (34.59 vs. 34.55 words, respectively, for Madison's and Hamilton) and the frequency of such telltale words as "upon" and "whilst" in Madison and Hamilton's prose. But in the end, they used such noncontextual words as "by" and "from" to show that Madison had written the 12 disputed essays.
Their analysis, published in their 1964 book "Inference and Disputed Authorship," spurred consensus among historians over their findings and was an early and persuasive demonstration of what has come to be called "stylometry."
Stephen E. Fienberg, a former doctoral student of Dr. Mosteller's and a statistics professor at Carnegie Mellon University in Pittsburgh, placed Dr. Mosteller among "a dozen real giants of the last century in statistics."
Fienberg said Dr. Mosteller's legacy was to show how adaptable many of his methodologies were to matters that were seemingly remote from one another.
He said the mathematical model Dr. Mosteller developed in the 1960s to look into surgical risks from the anesthetic Halothane, "for which there were many categorical variables and relatively sparse data," could be applied to such diverse problems as measuring the impact of human rights violations in Kosovo and assessing disability among the elderly.
Charles Frederick Mosteller was born Dec. 24, 1916, in Clarksburg, W.Va., and raised near Pittsburgh. He paid for his early college education at what is now Carnegie Mellon by working with his father, a road builder, and by winning poker games.
At college, he became fascinated by matters of probability and was smitten when a mathematics professor showed him sophisticated problem-solving techniques.
"It was the most marvelous thing I had ever seen in mathematics," Dr. Mosteller told the authors of the book "More Mathematical People."
"It used mathematics that, up to that time, in my heart of hearts, I had thought was something that mathematicians just did to create homework problems for innocent students in high school and college," he said.
Dr. Mosteller received a bachelor's degree in 1938 and a master's degree in mathematics in 1939, both from Carnegie Mellon. After receiving his doctorate in mathematics from Princeton University in 1946, he joined the Harvard faculty in the social relations department. He became the first chairman of Harvard's statistics department, from 1957 to 1971, and retired in 1987 as chairman of the department of health policy and management.
Dr. Mosteller reached a national audience in 1961 when he taught a course in probability and statistics that aired on NBC's early morning program "Continental Classroom."
He also was president of several professional associations, served on scientific research boards and continued an active research schedule until moving to Northern Virginia from Belmont, Mass., in 2004.
Among the works in his prolific output were several articles on magic tricks and bridge. One of Dr. Mosteller's early papers, the first known academic analysis of baseball, showed that even a strong team relies heavily on luck in a short, seven-game series. He wrote the piece after the Boston Red Sox, his favorite team, lost the World Series to the St. Louis Cardinals in 1946.
His wife of 60 years, Virginia Gilroy Mosteller, died in 2001.
Survivors include two children, William Mosteller of Falls Church and Gale Mosteller of Arlington County; and a grandson.