2/13/01

Dawn Levy, News Service (650) 725-1944; e-mail: dawnlevy@stanford.edu

Donoho tops list of most-cited mathematical scientists

Five Stanford statistics professors were among the top 15 mathematical scientists cited during the 1990s, according to Science Citation Index, a publication of the Institute for Scientific Information. David Donoho topped the list at 632 citations, followed by his frequent co-author Iain Johnstone (3rd with 559 citations), Brad Efron (11th with 382 citations), Jerome Friedman (12th with 373 citations) and Robert Tibshirani (14th with 356 citations). No other university's mathematics, statistics or applied mathematics department had more than two members listed.

A key reason the Stanford statisticians were cited so often was the fact that their work focused on useful methodology, Efron said: "One of the nice recent turns in the field of statistics, since 1970, is a greater interest in algorithms that help scientists with difficult data problems. This applies to most scientists these days, since the computer age has allowed much more ambitious data-collection methods. The Stanford department has been the world leader in this movement."

Joint appointments with other science departments helped foster the development of useful methodologies. Tibshirani, Johnstone and Efron have appointments in the Health Research and Policy Department of the School of Medicine, and Friedman holds a joint appointment at the Stanford Linear Accelerator Center.

To compile the list, editors of Science Citation Index first identified the 200 most-cited papers for each year in the 1990s. Then they counted individual citations for each of the authors of the 2,000 identified papers.

Donoho and Johnstone have worked together on a spectacularly successful series of papers concerning wavelets -- modern versions of the sine and cosine waves that mathematicians use to tease out the signal from the noise to reveal the structures of earthquakes, solar activity and other cyclical phenomena. The Donoho-Johnstone work helped make wavelets accessible for statistical analysis.

Efron, the Max H. Stein Professor, is the inventor of the "bootstrap," a general computer-based way of attaching plus-or-minus values to a statistical estimate (as in, for example, "57 percent of the public plus or minus 3 percent are in favor of subsidizing public utilities"). Efron and Tibshirani wrote a book on this topic in the '90s.

Friedman is co-inventor, with Stanford's Richard Olshen and the University of California-Berkeley's Charles Stone and Leo Breiman, of CART. CART is a popular way of generating the diagnostic trees used in medicine and many other disciplines.

Tibshirani and Stanford colleague Trevor Hastie invented a popular data-modeling method called "generalized additive modeling." They are currently working with Friedman on a book on modern methods for data mining. These are methods that can be applied to large complex data sets such as those in the Human Genome Project.

-30-