Statistics captures unpredictability of real world
Harvard’s small but active statistics department celebrated its 50th anniversary last week. There were two days of lectures and panels Oct. 26-27 at the Gutman Conference Center, and a noisy, social, and musical banquet at the Harvard Club of Boston.
The statistics faculty wants the rest of us to know how useful their science is in a world of proliferating information. Having a lot of data requires a scientific way to organize and analyze it to make informed decisions.
They also want us to know how far statistics has come since 1957 — that its methods are better than ever at capturing the complexity and unpredictability of the real world.
And they want us to know how widely employed statistics are at Harvard — in virtually every academic division, said Statistics Department Chair Xiao-Li Meng. He and his colleagues are assisting biologists, physicians, chemists, engineers — and even historians.
Meng likes to quote the late John Tukey, a statistics pioneer at Princeton: “We get to play in everybody’s backyard.”
On Feb. 12, 1957, the Faculty of Arts and Sciences formally voted to approve Harvard’s new Department of Statistics. There were just two minutes of discussion.
It was among the first such university departments, preceded by Stanford, Berkeley, and a few others. In the United States, the University of Iowa established the first college department of statistics — to analyze data about farm crops.
Since 1957, Harvard’s statistics faculty members have played an increasingly hub-like interdisciplinary role, collecting and interpreting data for a wide variety of disciplines.
With only 9.5 positions presently, Harvard’s statistics faculty also reached more than 1,200 undergraduates in a variety of courses last year, and projects higher figures for the future. They also teach or provide technical support in medicine, law, business, economics, and in the life, physical, and social sciences.
Every scholar collects information, and then has to make decisions. That means dealing with both the information and the uncertainties it engenders.
In turn, statisticians provide a mathematical language to deal with uncertainties, said Meng. They create models to uncover patterns in the data — “mathematical indicators of real phenomena,” he said.
Meng himself is providing statistical insight to projects on the regional effects of global warming, astrophysics, and visual signal processing. He’s also designing a way to inferentially estimate missing data in a large-scale mental health survey.
Academic statisticians elsewhere work on models that will help judge musical compositions, analyze literature, and simulate textures in video.
At Harvard, Samuel Kou, John L. Loeb Associate Professor of Natural Sciences in the Statistics Department, is helping investigate the biophysics of a single-molecule protein — breakthrough work featured in one of the conference’s dozen research posters.
At the celebratory conference, there were technical talks that looked at statistical inference, Bayesian computation, and the EM algorithm.
Other talks provided historical perspective on statistics since the days of the Eisenhower White House and Sputnik, when Harvard’s experts were holed up in the old Reliance Bank building on Dunster Street.
There were retrospectives on two of the original five faculty members who have since passed away, William Cochran and Frederick Mosteller. (It was Mosteller’s veiled threat to move on to the University of Chicago that prompted action on forming a department at Harvard.)
During the conference, there were also frequent expressions of how widely useful statistics is to both intellectual pursuits and practical applications.
Statistics scholars are at their best “when solving real-world problems,” said one-time federal litigator D. James Greiner II, who earned a Harvard Ph.D. in statistics this year, and is now an assistant professor of law at Harvard Law School and a Statistics Department affiliate.
“Statistics makes chemists better chemists,” he said. “It also makes lawyers better lawyers.”
Statistics can also make real contributions in the humanities.
In 1966, Mosteller and University of Chicago statistician David Wallace published a groundbreaking statistical study of literary provenance in the journal Biometrics. They used complex mathematical statistics and the theory of inference to investigate the disputed authorship of 12 of the 85 Revolutionary War-era Federalist Papers. (The real author, they proved, was James Madison.)
Harvard President Drew Faust, Lincoln Professor of History, visited the celebration the first morning. She called statistics “a very important milestone … in the history of ideas” — and one that allowed “the animating spirit of trespassing freely across intellectual boundaries.”
Faust’s sixth book, “This Republic of Suffering” — due out next year — describes how Americans coped with a death toll of 620,000 from the Civil War, 2 percent of the nation’s population. “They turned to numbers,” she said, “as a way of trying to understand.”
Attendees at the anniversary conference — about 240 — also heard from Jeremy Bloxham, dean for the physical sciences and acting dean for the life sciences at Harvard. (He is also Harvard College Professor, the Mallinckrodt Professor of Geophysics, and a professor of computational science at Harvard’s School of Engineering and Applied Sciences.)
Bloxham praised the Department for its “intellectual diversity,” and for being central to the “broad intellectual landscape” of the Faculty of Arts and Sciences.
Mosteller, who died last year, typified the polymaths statisticians can be. He applied statistical techniques to medicine, business, education, and many more fields of inquiry. After statistics, he went on to chair three other departments at Harvard, a feat unmatched since.
Mosteller “felt he could have a foothold in any arena,” said Massachusetts Institute of Technology computational scientist Emery Brown, a physician who earned a Ph.D. in statistics at Harvard in 1988 and now also teaches anesthesia at Massachusetts General Hospital, a Harvard Medical School affiliate. “Fred started [interdisciplinary work] before it was a buzzword.”
Today, statistics has an important role to play in shaping the new general education curriculum for undergraduates, said Bloxham. Students will leave Harvard not only with good qualitative critical skills, but “capable of good qualitative criticism,” he said — “an ability to see through [any] statistical arguments being made.”
Stephen Blyth agreed that is important. He earned a statistics Ph.D. in 1992 and is now a visiting scholar in the department, as well as vice president for the international fixed-income portfolio at the Harvard Management Company. To understand risk, said Blyth, one needs to be able to “reason about uncertainty.”
Most departments at Harvard want to double in size, said Bloxham. “In the case of statistics, I don’t think we need to think very hard about that,” he said. “It needs to grow, and grow very substantially.”