Stephen Fienberg: Applying Statistics to Make a Difference
Professor. Researcher. Writer. Editor. Stephen Fienberg has taken different roles throughout his decades-long statistics career. His focus is on developing methodologies for statisticians to use in other fields of study.
“What I like to do in lots of areas is develop methodology for people to use with the hope that it will be used … and make a difference, whether it’s a message for census-taking or counting casualties in human rights context or privacy protection and confidentiality, all of the different kinds of things that I do,” Fienberg said.
Fienberg has authored more than 20 books and 400 papers and related publications. He has served as the dean of the Carnegie Mellon University (CMU) College of Humanities and Social Sciences and vice president for academic affairs at York University, as well as on the faculties of the University of Minnesota and University of Chicago. He currently teaches statistics at CMU. Earlier this year, Fienberg was awarded a grant for a project titled “Data Integration, Online Data Collection, and Privacy Protection for the 2020 Census” from the National Science Foundation-Census Bureau Research Network.
Born and raised in Toronto, Fienberg attended the University of Toronto from 1960 to 1964. He was originally enrolled in honors mathematics, physics, and chemistry.
“People came in with an interest in that range and had the same first-year courses, and began to specialize after year one,” Fienberg explained. “So after the first year, I dropped chemistry; after the second year, I dropped physics … but I took my first course in statistics my third year.”
That first statistics course was taught by Don Fraser, who later became Fienberg’s longtime friend and mentor. Under Fraser’s tutelage, many University of Toronto students went on to make a name in statistics. Fraser recently earned the Order of Canada for his contributions to the field of statistics.
“I think it would be safe to say that, on the technical issues that he focuses on in his research, we’ve never seen eye to eye, but it never stopped us from having terrific interactions,” Fienberg said of his relationship with Fraser. “He has a geometric way of describing many of the concepts and things he was trying to get across and, as an undergraduate, I was very much taken with geometric ways of looking at different mathematical concepts. The mix of geometry with statistics was intriguing. That was on the theory side. What really turned me on in a very different kind of way was the notion that this was really a set of methodologies to be used. In my fourth year, I began to learn about how statistics had made a major impact in a number of different fields, and the prospect that I might contribute to something like that, rather than simply trying to do mathematics, was what made me want to go on to study statistics in graduate school.”
Fienberg earned his undergraduate degree in mathematics and statistics in 1964 and went on to Harvard University to study in the department of statistics, a “very small department” at the time.
“It was part of the golden age of that department, and there was a terrific group of fellow graduate students,” Fienberg recalled.
During his time at Harvard, Fienberg grew close to assistant professor Paul Holland, fellow graduate student Yvonne Bishop, and Fred Mosteller, founding chair of Harvard’s statistics department. Not only was Mosteller Fienberg’s thesis adviser, he became a mentor and friend.
“Fred was not just an adviser on my thesis, but on many, many things over the years, and we interacted in a variety of ways up until his death in 2006,” Fienberg explained.
Fienberg earned both his master’s degree in statistics in 1965 and his PhD in statistics in 1968 from Harvard.
While Fienberg was a graduate student, Mosteller was involved in research in the National Halothane Study. Halothane was used as an anesthetic, and there were several case studies published in medical journals about people who had operations in which halothane was used as an anesthetic and died subsequent to the operation. The cause of death was usually unrelated to anything the patients had suffered at the time of their operations. Mosteller and a group of statisticians, including Bishop, were involved in data collections and analysis from more than 25 hospitals across the country, comparing death rates from different anesthetics. That work, Fienberg said, led to what is known as modern literature on nonlinear emergent models for categorical data.
American Statistical Association Founders Award, 2009
Statistical Society of Canada’s Lise Manchester Award, 2008
American Statistical Association Wilks Award, 2002
COPSS Presidents’ Award – Outstanding Statistician Under 40, 1982
Fellow of the Royal Society of Canada
Fellow of the American Association for the Advancement of Science
Fellow of the American Statistical Association
Fellow of the American Academy of Political and Social Science
Member, U.S. National Academy of Sciences
“Over the next several years, Yvonne, Paul, and I worked on different aspects of this and ultimately were the authors of a book called Discrete Multivariate Analysis, published by MIT Press in 1975, in which we laid out what was the basic theory of the time and some of the mathematical underpinnings,” Fienberg recalled. “There were lots and lots of applications and extensions. It included a lot of work from students from the department.”
It was through Mosteller that Fienberg met fellow statistician Judy Tanur during the early 1970s. “Fred was an organizer for many, many different things, but one of the activities he led around 1970 was something called the ASA-NCTM (National Council of Teachers of Mathematics), and they were set up to produce educational materials,” Fienberg explained. “They set out to develop two different products. One ended up being called Statistics by Example; the second was called Statistics: A Guide to the Unknown. Fred organized a meeting in Cambridge to which I was invited and Judy was there, and she ended up being the editor for the second project. That’s when we first interacted. I did do several pieces for Statistics by Example.”
Tanur and Fienberg didn’t interact again until the ASA meeting in Boston during the summer of 1976. Fienberg, Bishop, and Holland were teaching a continuing education course based on their book Exploratory Data Analysis. Tanur attended the course, and she and Fienberg started to talk.
“We became quite friendly in discussions and had gone out to dinner and then got together with Fred at some point during the meetings one of the evenings,” Fienberg recalled. “We then interacted in a number of different ways. Beginning in 1978, we began to go and visit as a family in Montauk, where Judy and [her husband] Mike had a small cottage. My wife and I went with our children and our dog every summer until the kids were grown. And Judy and I would write papers and develop different project ideas, and much of our collaboration was work that was done out at the deck looking out south toward the ocean.”
Fienberg still visits Tanur at her home in Montauk. Fienberg and Tanur co-edited Mosteller’s autobiography, The Pleasures of Statistics: The Autobiography of Frederick Mosteller, which was released in 2010.
Carnegie Mellon: A Place for Exploration
Fienberg began teaching at Carnegie Mellon University in 1980 and has remained there. “Carnegie Mellon is a very special place,” Fienberg said. “Most people talk about places where people do interdisciplinary research. We just do it.”
While at Carnegie Mellon, Fienberg has collaborated with a group of professors who worked with the computer science department to form the department of machine learning, as well as in the Heinz College and a variety of IT fields.
“The thing about CMU is we really did collaborate and cross boundaries with very little obstacles,” Fienberg said. “In my case, I’ve worked with people in a dozen different units.”
Fienberg currently serves as the Maurice Falk University Professor in the department of statistics. He also teaches in the machine learning department, Cylab, and i-Lab.
Fienberg’s attitude about statistics education has changed since he began teaching more than 30 years ago. “I think it needs to be taught well and taught by people who understand the details of what they’re teaching,” he explained. “In the past, and still, it’s been taught by people who either have only a rudimentary notion of statistics, or, often, too narrow perspectives. But there are lots of wonderful scientists out there who are every bit as knowledgeable about important aspects of statistics as are people trained with PhDs in statistics.”
Polygraph Accuracy Study
Fienberg has been involved in committees at the National Research Council for 35 years. “They are very fertile grounds for discovering new places to bring statistics,” he explained.
At one time, he served on the Divisional Advisory Board for the Division of Behavioral and Social Sciences and Education. The executive director asked Fienberg to serve as chair of a new committee she was setting up and to participate on a committee that was exploring technical aspects of a study done on polygraph accuracy.
“I had two principal qualifications,” Fienberg recalled. “One was I had never read any literature on the polygraph and, two, I had never had one. So I brought no direct biases to the table when I had to evaluate what the literature had to say about polygraph accuracy.”
The committee was a consequence of the Wendell Lee case, Fienberg explained. Lee was accused of being a spy for the Chinese, and Congress told the Department of Energy to attend the National Academies and have them perform a study on the accuracy of the polygraph for security purposes.
The committee concluded “government agencies could not justify their reliance on the polygraph for security screening.” The DOE’s proposed regulations disregarded the committee’s findings and conclusions.
“We had been what I would say substantially negative about the use of polygraphs for security screening, but they intended on going ahead and doing what they had been doing,” Fienberg said. “At which point, the ranking members of the Senate oversight committee set up the hearing.”
Fienberg enjoys applying statistics to different fields to help others reach their goals, whether in research or public policy.
“One of the things Fred Mosteller taught me that has affected, I hope, almost everything I do is that [statisticians] are not supposed to be negative, telling people they shouldn’t do something or what they did was no good,” Fienberg said. “Our role is to help people do good things with statistics that will address their scientific or policy goals and also … technical goals for the field.”
Fienberg currently resides in Pittsburgh with his wife. He plays hockey in his spare time, an activity he says is a “wonderful way for me to try to blow off steam and keep a little bit fit.” He also enjoys spending quality time with his grandchildren.