Thursday, October 23, 2008

David A Freedman

BERKELEY: ­ David A. Freedman, a professor of statistics at the University of California, Berkeley, who fought for three decades to keep the United States census on a firm statistical foundation, died Friday, Oct. 17, of bone cancer at his home in Berkeley. He was 70.

Throughout his career, Freedman made major contributions to the theory and teaching of statistics. But he also had a broad impact on the application of statistics to important medical, social, legal and public policy issues, including clinical drug trials, epidemiologic studies, economic models, interpretation of scientific experiments, statistical evidence in the courtroom and adjustments to the census.

"David transformed the practice of applied statistics as it is directed toward litigation, toward Congressional action and toward public policy," said long-time friend and colleague Kenneth Wachter, UC Berkeley professor of demography and statistics. "The prevailing mode when he began working was to rely on hypothetical models with assumptions sometimes driven by mathematical convenience, which were fine for theoretic work but, when carried over to applications in the policy arena, gave conclusions that were often fanciful or driven by the prejudices or presuppositions of the statisticians testifying or contributing."

"Not only has David, since his early twenties, been recognized as one of the world's leading mathematical statisticians, but he has also assumed the mantle as the skeptical conscience of statistics as it is applied to important scientific, policy and legal issues," wrote James M. Robins, professor of epidemiology at the Harvard School of Public Health, in 2002.

Freedman clarified the assumptions underlying a wide variety of statistical models and revealed how sensitive conclusions can be to violations of the those assumptions - regardless of the quality of the data. "By distinguishing proposals based on hypothetical modeling from proposals grounded in empirically established observations, he developed a firmer basis for applying statistics to policy," Wachter said.

His legacy, said UC Berkeley colleague Philip Stark, professor of statistics, is "demystifying and debunking many of the tools people use in social science and elsewhere to try to draw inferences." Even today, "there is a lot of muddled thinking and blind reliance on methodology - almost a religious belief that methods give truth - without looking carefully at the assumptions of the methodology. David contributed enormously to the clarity and rigor and circumspection" in the field of applied statistics.

"My strongest childhood memories of my father are sitting at the kitchen table, and David recounting some study (and) then delightedly challenging me to see the flaws in the design," said his son, Joshua Freedman. "He had a driving passion for rigor and the glimmers of truth found in good data."

Both Freedman and Wachter testified before Congress and the courts against adjustments to the 1980 and 1990 censuses proposed to make up for perceived geographical and ethnic undercounts. A 1990 lawsuit to force the Department of Commerce, which oversees the decennial census, to make such adjustments was taken all the way to the U.S. Supreme Court, which in 1996 sided unanimously with the Commerce Department and Freedman's analysis. The department won a similar lawsuit in 1980.

"The census turns out to be remarkably good, despite the generally bad press reviews," Freedman and Wachter wrote in a 2001 paper published in the journal Society. "Statistical adjustment is unlikely to improve the accuracy, because adjustment can easily put in more error than it takes out."

Freedman wrote a widely used reference guide on statistics in the courtroom published by the Federal Judicial Center, the education and research agency of the federal courts, and was viewed by many as the world's leading forensic statistician, Stark said. Freedman testified as an expert witness on statistics in law cases that involved employment discrimination, fair loan practices, voting rights, duplicate signatures on petitions, railroad taxation, ecological inference, flight patterns of golf balls, price scanner errors and sampling techniques. He worked as a consultant for the Carnegie Commission, the City of San Francisco and the Federal Reserve, as well as the U. S. departments of energy, treasury, justice and commerce. He was often called by the media to comment on the statistical validity of studies.

In 2003, Freedman was awarded the prestigious John J. Carty Award for the Advancement of Science by the National Academy of Sciences, "for his profound contributions to the theory and practice of statistics, including rigorous foundations for Bayesian inference and trenchant analysis of census adjustment."

Freedman was deeply committed to improving the quality of statistics education, said colleague David Collier, UC Berkeley professor of political science. As chair of UC Berkeley's statistics department from 1981 to 1986, Freedman and his colleague Peter Bickel reorganized undergraduate teaching to emphasize the applied aspects of statistics. Freedman regularly taught a graduate course in statistical consulting and supervised the Statistical Consulting Service, which served campus researchers in a broad spectrum of disciplines while providing real-world experience for statistics graduate students.

"Freedman's transition from being a mathematical statistician to a creative practitioner of applied statistics occurred in part, by his own account, in response to the challenges of undergraduate teaching on the UC Berkeley campus," Collier said. "His students were bored with statistics courses and with the abstracted examples that were standard fare in textbooks," leading Freedman to dig up practical examples in many applied areas.

Wachter noted that, thanks to the late Jerzy Neyman, who founded the field of modern statistics and UC Berkeley's statistics department, "Berkeley was famous for statistical theory. If you wanted to do theory, Berkeley was the place, whereas applied work was given short shrift. David and Peter Bickel undertook transforming the department as the wave of theory had run its course to emphasize high level, mathematically informed applied work, since significant theoretical work now comes in response to applied problems."

He wrote six textbooks, including the highly regarded undergraduate text, "Statistics," with co-authors Robert Pisani and Roger Purves, now in its fourth edition. The book is "a widely used undergraduate textbook, crystal clear, a delight to read and to teach from, broad, deep and meticulously accurate in every detail," Stark said. "It transformed the way many people taught statistics from a formula-driven, plug-in-the-numbers approach to a focus on critical thinking."

Throughout his career, Freedman sustained a deep intellectual and personal engagement with a wide circle of colleagues and students, Collier said. "Given his stunning insights, his thoughtful advice, and his astute wit, exchanging ideas with him was a memorable source of learning, and of fun."

Freedman was born in Montreal, Canada, on March 5, 1938, and obtained his B. Sc. from McGill University in 1958 and his Ph. D. from Princeton University in 1960. After a year at Imperial College London on a Canada Council fellowship, he joined the UC Berkeley statistics department in 1961 as a lecturer, and was appointed to the faculty in 1962. In addition to stints as vice chair and chair of the statistics department, he also was a Miller Fellow in 1990 and an Alfred Sloan Fellow in 1964-66. He was a member of the American Academy of Arts and Sciences.

"It was evident early on that David had a much broader range of interests and talents than a typical academic statistician," Robins noted. "After doing groundbreaking research on nonparametric Bayesian estimation, Markov chains, and other areas of mathematical statistics, David took a two-year leave of absence to work for the Bank of Canada to learn and contribute to the application of statistics to economic analysis."

He subsequently immersed himself in analyses of econometric models and financial issues, and also came to focus on the field of epidemiology, where he critiqued or consulted on studies of lung cancer, heart disease and mad cow disease, among others. He published much of this analysis in his 200 papers and 20 technical reports.

"The fundamental importance of David's contribution to the application of statistics is evident when one recalls that today statistically illiterate researchers have available, at their finger tips, high-powered computer packages that perform multivariate analyses and regressions that seem to provide magical black-box answers to scientific questions," Robins wrote. "It was essential that someone in the statistical community stand up and say that statistical models and regression are not a magic cure for inadequate data, plagued by uncontrolled confounding and measurement error."

Freedman is survived by his wife, Janet Macher; stepmother, Charlotte Freedman of Montreal, Canada; two children, Joshua of Corralitos, Calif., and Deborah Freedman Lustig of Walnut Creek, Calif., his first wife, Shanna Helen (Wittenberg) Swan of Rochester, N.Y.; and four grandchildren.

A campus memorial is planned for December. In lieu of flowers, donations in memory of David Freedman can be made to the UC Berkeley Foundation, c/o University Relations, 2080 Addison St., Berkeley, CA 94720-4200.

No comments:

Post a Comment