At a meeting of the Faculty of Arts and Sciences on Oct. 3, 2023, the following tribute to the life and service of the late Carl Neracher Morris was spread upon the permanent records of the Faculty.
Carl Morris was one of the most accomplished statisticians of his generation: his contributions ranged from mathematics to health care policy to sports statistics. Morris held appointments at the University of California, Santa Cruz, the RAND Corporation, the University of Texas at Austin, and the Harvard Departments of Health Care Policy and of Statistics. He was a Fellow of the American Statistical Association, the Institute of Mathematical Statistics, and the Royal Statistical Society, and was an elected member of the International Statistical Institute. He served as the executive editor for Statistical Science and as editor of Theory and Methods for the Journal of the American Statistical Association.
His dedication as a teacher extended beyond the classroom: he was a valued student advisor and a supportive mentor for junior faculty. Many statisticians relied on his fundamental work in statistical theory as well as his generous, expert counsel. His wry sense of humor delighted all who knew him and could transform a dull class or meeting into something unexpectedly delightful. Conversations with Morris were often surprising, always pleasantly so. He might be describing the baseball pennant race, upcoming tennis matches, some deep connection between probability distributions, or insights on teaching.
Morris’s contribution to statistical theory began with his co-author and lifelong friend, Brad Efron. Morris and Efron had been undergraduates together at Caltech and, later, graduate students at Stanford. In a seminal series of papers in the 1970s, Morris and Efron showed how seemingly disparate ideas could be unified into an elegant whole and, along the way, demystified a conundrum in statistical theory — Stein’s paradox. Their papers showed that prediction could be improved using Bayesian models with prior distributions estimated from data and that the James-Stein shrinkage estimator, which had shocked the field of statistics by showing that predictions could be improved by augmenting data with seemingly unrelated observations, was a natural consequence of their approach. This estimator was shown by their work to be useful in applied statistics rather than being simply a theoretical curiosity. The papers are widely considered the foundation of modern empirical Bayes, were widely used before powerful computing provided other approaches, and are still used today in some deep learning algorithms. Morris’s eclectic interests were evident even in this theoretical work. Efron and Morris showed that end-of-season batting averages for a set of major league baseball players were substantially more accurately predicted when early season averages for each player were replaced by the James-Stein estimator using data from all players in their dataset. In the words of Andrew Gelman at Morris’s retirement celebration, Efron and Morris “de-paradoxed” Stein’s paradox.
Morris’s long-admired intellectual generosity was evident in his earliest work. When asked in an interview for the publication CHANCE about the idea of using baseball statistics to illustrate empirical Bayes, Morris said that he and his co-author Brad Efron “both admired Charles Stein so much for his genius and his humanity, we chose this topic, hoping we could honor him by showing that his estimator could work well with real data.”
Morris and Efron did not simply stumble on the batting average data: Morris had long been a certifiable baseball statistics fan(atic). In fact, after studying aeronautical engineering at Caltech, Morris figured that, by studying statistics, he might land his dream job — statistician for one of the major league baseball teams. His love of baseball began in his youth. There were no major league baseball teams west of Kansas City so, in addition to watching minor league teams in the Pacific Coast League, Morris followed baseball through AM radio and morning box scores in newspapers. Many statisticians are baseball fans, some even dabble in baseball statistics, but Morris was not a dilettante. Two of us (David Harrington and Joseph Newhouse) spent hundreds of hours as youths playing the spinner-based game All-Star Baseball but never tailor-made disks, as Morris did. He later used multilevel modeling to assess whether Ty Cobb was truly a .400 hitter or whether he was only lucky in a few seasons. Yes, he likely was a .400 hitter, concluded Morris. At Harvard, Morris became the founding faculty advisor for the Harvard Sports Analysis Collective (HSAC), where his steady presence and generous counsel inspired students to begin careers in sports analytics. As HSAC noted on its website, “The club would not exist in its current form without his dedication, mentorship, and support.” Morris regularly attended meetings, wrote papers with student members, and advised senior theses.
In addition to his work on empirical Bayes, Morris made important contributions to probability theory. In a paper considered a breakthrough in statistics, he showed that there are exactly six families of probability distributions that are members of the natural exponential family and have a quadratic variance function. Five were well known — Morris discovered the sixth to complete the set and derived many special properties enjoyed by these families. He made contributions to the hierarchical models widely used today in the social sciences, economics, and health care policy. As the lead statistician for RAND’s famous Health Insurance Experiment, Morris devised a new randomization scheme, the Finite Selection Model, that was more efficient than standard models and important to the study’s success.
Morris held important leadership positions, most notably at Harvard where he chaired the Department of Statistics from 1994–2000, years that were not without controversy. The department was small but had internationally recognized scholars and a vibrant graduate program. The undergraduate concentration was struggling, however, and had never had more than 10 total concentrators. There were arguments within and outside the department for dropping the statistics concentration. Morris recognized the importance of statistics in shaping undergraduates’ understanding of the world beyond Harvard. Despite challenges, he believed in sustaining the concentration during tough times. Thanks to Morris’s vision, the department now boasts internationally acclaimed programs and educates many Harvard students, including approximately 200 current concentrators.
Morris was a loving father and grandfather; he is survived by his three children and five grandchildren. He is sorely missed by family, friends, and colleagues.
David Harrington, Chair