Gong Tang, Associate Professor of Biostatistics at the University of Pittsburgh, and Michael R. Elliott, Professor of Biostatistics and Research Professor of Survey Methodology at the University of Michigan

From left: Gong Tang, Rod Little, and Michael R. Elliot at the UM Department of Biostatistics in 2015

Student Years

Tang: Rod, you had a bachelor’s degree in math. What made you decide to pursue graduate study in statistics?

Little: The career advice for mathematicians at Cambridge at that time favored operational research (OR), and when a job didn’t materialize for me I was looking to do a master’s degree in that field. There was a master’s program in statistics and OR at Imperial College, and I applied because of the OR part, and because my girlfriend at the time happened to be in London. The program turned out to be 90% statistics and 10% operational research, so I ended up in statistics. It turned out to be a wonderful move.

Rod Little gives the Fisher Lecture in 2012.

Tang: You finished your master’s and doctoral degrees in statistics within three years. How did you finish so efficiently?

Little: It was not unusual in England at that time. One had a three-year grant, and was expected to complete in that period. I was fortunate to have a very supportive adviser, Sir David Cox. My unofficial advisor was Martin Beale, with whom I ended up working. I was also lucky to have a fruitful research topic.

Early Career

Elliott: In 1974, you started your career as a faculty member at the prominent department of statistics at the University of Chicago. After two years, you went back to London to work at World Fertility Survey (WFS) for the International Statistical Institute. How did you make those career choices?

Little: The statistics department at Chicago was looking for people who could help with consulting, and they had the strange idea that English people could analyze data. They contacted Sir David Cox, and he gauged my interest. It was a great opportunity, even though the position was short term. During the second year, Sir Maurice Kendall came through to recruit people to work on this fantastic project that would make everyone famous, the WFS. He encouraged me to apply, and it turned out to be a wonderful way to start a statistics career.

Elliott: What was good about the WFS?

Little: It was a real study, the largest social science project in the world at that time. There was a smart young group of demographers, statisticians, and computer scientists, led by a famous and charismatic statistician, Sir Maurice Kendall. I learned a lot from the demographic application about connecting statistics and the real world. It was not too onerous, I had the flexibility to write my own research, and I got to travel and teach widely in developing countries.

Elliott: In 1976, Don Rubin’s landmark paper “Inference and Missing Data” was published with your formal comment in Biometrika. You and Don later became close collaborators. How should young researchers think about collaborations and choosing good collaborators?

Little: I have had the good fortune to learn from and work with a number of top-notch statisticians—Sir David Cox, Don Rubin, and Sir Maurice Kendall, to name just three. If you have a chance to work with one, you should jump on the opportunity, even if other jobs are better paying or have more prestigious titles.

Elliott: those early years of your career, you published a few papers about missing data and sample surveys in journals such as JASA and the Journal of the Royal Statistical Society, Series B. Was the research motivated by your work at the WFS?

Little:The survey part, yes—missing data were not as serious an issue in the WFS because people in developing countries generally responded. Leslie Kish was involved in the designing of surveys, and there was a big survey sampling component. There was also a modeling component because the demographers used models. So I got interested in the design-based versus model-based approaches to survey data analysis.

Years at UCLA

Tang: In 1983, you moved to the department of biomathematics in the school of medicine at the University of California at Los Angeles. Can you shed some light on the move and the transition to include medicine in your fields of application?

Little: After the WFS, I spent three formative years working with Don Rubin at the Environmental Protection Agency (EPA), and then as a fellow at the Census Bureau in Washington, DC. Those jobs were great but short-lived. Wil Dixon, a well-known statistician who started the statistical software BMDP, was at UCLA and formed a good opinion about me from my adviser Martin Beale. He pushed strongly for my appointment as tenured associate professor in the department of biomathematics, although I had not previously held a tenure-track position. I did get another offer from University of Washington at Seattle for a joint appointment between statistics and sociology. But the future funding for that position was uncertain. I decided to go to UCLA because it was tenured. (The UW job was eventually taken by Adrian Raftery, so I think of Adrian as my gift to that university.) The biomathematics program at UCLA was small, but distinguished. I was hired for the statistical component and got more involved in biomedical consulting as part of the switch from statistics to biostatistics.

Rod Little sings at a University Musical Society Choral Union rehearsal.

Tang: During the next 10 years at UCLA, you were principal investigator on one National Science Foundation (NSF) grant and two National Institute of Mental Health (NIMH) R01 grants. What are important factors for success in securing research grants, especially for young investigators?

Little: It was a little easier than nowadays. I think one needs to have good ideas, a good publication record, and the ability to write well. You can’t win unless you apply, and try to learn from rejections. It is tough but a good way to force yourself to think about what you are doing and where you are going, even if you don’t get funded. Persistence is important.

Tang: During your years at UCLA, you published several influential works on missing data and sample surveys, and established your status in these fields. You also published a classical textbook, Statistical Analysis with Missing Data, with Don Rubin in 1987. What stimulated such productivity?

Little: The book with Don Rubin actually started earlier when I was at the EPA—books take a while to complete. The job was mainly doing research, writing papers, and teaching, and I guess I tried to do it well. I was fortunate to have time to teach, do methods research, and work on that collaboration.

Years at Michigan

Elliott: In 1993, you became the chair of the department of biostatistics at the University of Michigan. Was it primarily because you were interested in leading an academic program, or were you also attracted to Michigan for other reasons?

Little: It was a combination. I enjoyed Los Angeles, but my wife, Robin, was not so keen on the earthquakes, pollution, and life in the big city. I was attracted to Michigan because of its strong group of faculty and graduate students—I felt there was great potential there. I was acting chair of biomathematics at UCLA for a year, so I got my feet wet with administration. As you know, Ann Arbor is a great town, and Robin and I never regretted the decision.

Elliott: Besides continuing research in missing data and sample surveys, you started expanding your research in causal inference as well as Bayesian statistics. Can you share with us some thoughts about how you kept developing new research topics and making fresh contributions?

Little: At the WFS, Sir Maurice Kendall was interested in how to do causal inference in observational studies, and in particular path analysis. Rubin worked on causal inference and stimulated my interest because he viewed causal inference as a missing-data problem. I always have had a pretty broad range of statistical interests, partly because missing data arise in many areas of statistics, and work in missing data led me to consider various areas. Maybe I picked up my taste for broadness from my father, who was a journalist with very eclectic interests. Anyway, I guess my contributions tend to be broad rather than specialized. Others focus on a narrow area and make deep contributions, and that’s obviously fine too. My advice is to just do what you do best.

Elliott: When you were chair, you recruited a number of junior faculty members who later became highly successful statisticians in their own right—including Robert Strawderman, Xihong Lin, and Debashis Ghosh—who are now department chairs. They had different strengths and quite different research agendas. How did you help them to start their career?

Little: By leaving them alone to do their own thing, while trying to be as supportive as possible. I did not come to Michigan biostatistics with a narrow research agenda for the department, and see the wide range of research interests as a big strength.

Elliott: Since 1993, the department of biostatistics at the University of Michigan has grown dramatically, producing many outstanding graduates and becoming one of the top graduate programs in biostatistics. What advice would you give a new chair to develop a successful graduate program in biostatistics?

Little teaching online Biostatistics.

Little: I had a supportive dean who valued statistics as more than a service discipline. Recruit the best students, staff, and faculty that you can—it’s all about the people. We try to nurture a democratic, collegial atmosphere at Michigan, and that helps to recruit good people. As chair you have the responsibility to run the department well, seek input but make tough decisions when needed, and support your colleagues. It’s not about you.

Elliott: From 2010– 2013, you served as the inaugural associate director for research and methodology and chief scientist at the U.S. Census Bureau. What motivated you to take that challenge?

Little: Bob Groves, who was the head of the Survey Research Center at Michigan, was appointed as the director of the Census Bureau. He previously had a research role in the census and was interested in creating a new research and methodology directorate, to give research more visibility. He knew me from working together on missing data and surveys, and asked me to be the first leader of the new directorate. It was a good opportunity to make an impact on government statistics, which has been an interest of mine. Both Robin and the university were supportive. It was an interesting couple of years. The new directorate was formed, and I believe is playing a pivotal research role in bureau activities. The leadership is rotating under Groves’ vision, and I have great successors in Tom Louis, and the person taking over from him, whom I know but is yet to be announced.

Service and Others

Tang: Mentoring students has been an important part of your career. Do you have any broad guidelines you could share about this process?

Little: I think there is a wide range of styles and many can work. My approach also depends on the student, so it’s not one size fits all. I try not to be too intrusive or put too much pressure on students, since I think they are usually motivated to succeed without my nagging. The initial research ideas generally come from the adviser, and the objective is to have the student’s contributions toward the research grow over time as she or he becomes more mature and understands the field better. I have had great students, and letting them flourish is the main thing.

Tang: You published numerous papers in top journals such as JASA, JRSSB, Biometrika, and Biometrics and served in various editorial roles. Do you have any tips for publishing in and editing for top journals?

Little: Statistical research varies from theoretical to applied, but being a good writer improves the chance of good publications regardless of where you are on the spectrum. So time spent improving writing is well spent. I have style tips on my web site. I benefited from being the son of a journalist, who was interested in writing clearly and concisely. I see no harm in trying top journals, aside from some damage to one’s ego. You will probably get rejected the first few times, but you will learn from the reviewers. Editing also involves good writing, and is a kind of learn-as-you-go process. You have to get on the bandwagon by letting people know that you are interested in refereeing papers. If you do a good job refereeing, you are likely to be invited to be an associate editor down the line.

Tang: You have served as vice president of the ASA and on many national committees. What are the joys of serving those committees?

Little: Membership organizations such as the ASA, IBS, ENAR, and WNAR have been a great help to me in my career. I encourage all professional statisticians—in academia, government, or industry—to participate in these societies. As for being ASA vice president, the ASA staff and board members are great, and it is fun to work with smart lively people. In particular, ASA Executive Director Ron Wasserstein has been fantastic and is a pleasure to work with.

Tang: Besides your work, you also enjoy time with your wonderful family and participate in community activities—including a significant amateur singing career and a leadership role on the board of the Ann Arbor Symphony Orchestra. How do you find that balance between work and other components of life?

Little: As the old saying goes, “All work and no play makes Jack a dull boy.” Doing these other things keeps me happy and more productive. I think people should work hard and they should play hard as well.

Philosophy of Statistics

Elliott: In the book This Idea Must Die: Scientific Theories That Are Blocking Progress, Emanual Derman says “Statistics—the field itself—is a kind of Caliban, sired somewhere on an island in the region between mathematics and the natural sciences.” Could you say a bit about how you see statistics fitting into the scientific endeavor, especially given the rise of “data science”?

Little: Ariel sounds more appealing than Caliban! I touched on this in my Fisher Lecture published in JASA a couple of years ago. I started in pure mathematics as an undergraduate, and clearly mathematics is an indispensable basis of statistics. Over time I got much more captivated by the scientific roots and applications of statistics, as in the work of Ronald Fisher or George Box. Statistics is not mathematics, its mathematical basis notwithstanding. The great attraction of statistics is that it can be applied in lots of different areas, and I have had the pleasure of doing this. To be sure I respect people who go deeply into the mathematical roots of the subject, but that is not one of my strengths.

Elliott: What has stayed constant in statistics over the past 40 years and what has changed? What do you think are likely to be among the most productive lines of research in statistical methodology over the next decade or two?

Little: I am a fan of a “calibrated Bayes” approach to inference—as espoused by Rubin, Box, and others—and have been advocating this approach for survey sampling, where the design-based approach is less and less compelling, given the growth of Big Data and challenges of obtaining true probability samples. More generally, I think the ideas of statistical theory, having good confidence coverage and the like, and the need for finding useful statistical models for a variety of problems is constant. I think statistics brings rigor to areas like Big Data, where good statistical modeling is constantly needed. What has changed is the enormous development of computational ability and tools. In particular, Bayesian statistics was not very feasible except in relatively small problems when I was a student. The enormous growth of computational power and accessibility to data makes statistics an exciting field to work in these days.

Writing a paper? Review Little’s style and grammar tips for biostatistics and statistics students.

This conversation occurred at Ann Arbor, Michigan, on April 28, 2015.