ORIE marks Bruce Turnbull’s retirement with a symposium and dinner
“Turnbull TakeOff!” honors ORIE’s 1971 PhD graduate and professor for 40 years
Distinguished academic and industry statisticians spoke at a symposium celebrating the retirement of Professor Bruce Turnbull, PhD ’71. They joined colleagues, students, family and friends at a barbeque, catered by Syracuse favorite Dinosaur Barbeque, at Cornell’s A.D. White House afterwards.
The gathering included the presentation of a slide show featuring Turnbull, family and colleagues over the years.
Turnbull, who first came to Cornell as a graduate student following a B.A. in mathematics from Cambridge University, has worked on a wide variety of statistical applications in energy, epidemiology, medicine, engineering, manufacturing and toxicology. His Cornell dissertation, under the direction of former ORIE Professors Howard M. Taylor III and the late Robert E. Bechhofer, is titled Bounds and Optimal Strategies for Stochastic Systems. Taylor, who has retired from the University of Delaware, attended the event.
Turnbull joined the Cornell faculty in 1976 following positions at Stanford and Oxford. At the beginning of this century he served as the first chair of the newly formed Department of Statistical Science and is retiring as a professor in both ORIE and that department.
Taylor recalled that there had been efforts to form a university-wide statistics unit in the late 196’s, when he was on the ORIE faculty, but there was much resistance to such a move at that time. “At Bruce’s retirement,” he said, “I was amazed at what he had achieved in terms of statistics at Cornell, where a university-wide Department now exists. What an achievement! Bruce can get things done – I think part of it is that he is a good listener.”
In recent years Turnbull’s research has focused on the design, monitoring and analysis of clinical trials that can adapt while in progress, based on one or more interim analyses, without compromising the statistical properties of the trial. Statisticians play a key role in planning the essential elements- such as the number of individuals to participate and how they are selected - of trials intended to determine whether the outcome of a medical intervention, such as a treatment or medication, is safe and effective. Turnbull’s research provides methods to guide the formulation of statistically sound designs.
As Turnbull and coauthor Christopher Jennison Ph.D. ’82 point out in their 1999 book on sequential approaches to experimentation, these approaches may date from Noah, “who on successive days released a dove from the Ark in order to test for the presence of dry land during the subsidence of the Flood.” Much more recently, the results of a sequential approach to trials can be seen in decisions to stop a trial for either positive or negative reasons, although as Turnbull and Jennison note, “the decision to stop the experiment is not primarily statistical.“ Nonetheless, “nearly all major long-term clinical trials now contain statements” describing procedures “for statistical monitoring of accruing results.”
The Turnbull Takeoff Symposium featured four statisticians with whom Turnbull has been associated during his career. Jennison, who carried out his Ph.D. research with Turnbull and has been a continuing collaborator, is professor and chair of mathematics and formerly Dean of Science at the University of Bath, England. David Hall Ph.D. ’83 CALS, took courses with Turnbull while a graduate student in plant breeding at Cornell and is now a Distinguished Statistician at Boehringer Ingelheim Pharmaceuticals in Connecticut, where Turnbull has served as a consultant. Keaven Anderson, now executive director of late development statistics at Merck Research Laboratories in West Point, Pa., met Turnbull when Anderson was a Stanford graduate intern at the National Laboratories in Oak Ridge Tenn. and Turnbull was on sabbatical leave there. And Turnbull has sailed on Lake Minnetonka Minn. with Tom Louis, professor of biostatistics at The Johns Hopkins Bloomberg School of Public Health, whom he has known both professionally and socially.
Nearly half of Turnbull’s Ph.D. students in operations research and statistics, including Jennison, attended the event, as did several Cornell Ph.D.’s who took courses with him. The earliest Turnbull-advised graduate is Russell Barton Ph.D. ‘78, who is now senior associate dean and professor of supply chain and information systems and industrial engineering in the Smeal College of Business at Penn State. Another advisee, Marc Meketon Ph.D. ’80, is a vice president at consulting firm Oliver Wyman, where he focuses on R&D for freight railway systems. Advisee David Amato Ph.D. ’82 is vice president for biometrics at Vertex Pharmaceuticals. Others include Lisa McShane Ph.D. ’89, chief of the biostatistics branch in the Biometric Research Program at the National Cancer Institute; Haiqun Lin Ph.D. ‘00, who came to Cornell with an M.D. from Beijing Medical University and is an associate professor of public health at the Yale School of Public Health, and Catalina Stefanescu-Cuntze Ph.D. '02, Dean of the Faculty and first holder of the Deutsche Post DHL Chair at the European School of Management and Technology in Berlin, Germany.
Dr. McShane recalled that when she was a student “Bruce insisted that I learn something about the science behind each of the projects that I worked on as his research assistant, and which later formed the basis for my thesis.” For example, in assessing the performance of a high-performance liquid chromatography (HPLC) assay system for measuring vitamin E in human blood plasma, “he had me visit the lab to learn how the samples were processed and see how the HPLC equipment was configured. The skills of engaging in the science and communicating with non-statisticians have served me well in my 25+ years as a statistician at the National Institutes of Health, including the last 20 in the National Cancer Institute where I have worked extensively on studying molecular profiles of tumors to predict prognosis and optimize therapy.”
Dr. Meketon described attending Turnbull’s seminar course as a graduate student, where he was exposed to “exciting new research,” including” a new-fangled non-parametric ‘double censored’ Maximum Likelihood Estimator (invented by Bruce)”. Although he has not done much work in statistics in his career, Meketon said that recent encounters reaffirmed his appreciation of Turnbull’s enduring influence: a statistician told him that “she needed the “Turnbull estimator,” and another, “apparently a bigwig at a pharmaceutical company, told me he hoped Bruce will return soon to help him out.” “I was fortunate to have Bruce as my advisor,” said Meketon. “He was helpful in many ways, and was very flexible as my research interests changed from one area of mathematical statistics to another.” In 2007, part of the thesis Meketon wrote under Turnbull’s guidance won an important award in the field of statistical analysis of simulation outputs.
Professor Lin said that “during the time as a graduate student studying with Bruce, I wrote my first paper for publication. Through the many rounds of editing guided by Bruce I discovered the highest standard of academic rigor that he follows.”
In his talk, introduced by Turnbull’s colleague Leon Welch Professor Emeritus Michael Todd, Professor Jennison reminisced about his first encounter with Turnbull exactly 37 years earlier in Turnbull’s graduate data analysis course – and showed a page from his course notes that verifies the date. He noted that over the years the Jennison and Turnbull families have grown close. Jennison then described their ongoing professional collaboration, which deals with the design of clinical trials. Their methods overcome ethical and practical difficulties clinical trials that may terminate when early results demonstrate marked success or failure, potentially disrupting the statistical properties on which the trial designs were based. The essence of their work is captured on the cover of their now-standard text Group Sequential Methods with Applications to Clinical Trials. (The cover illustration, reproduced on the commemorative glass at left, is due to their friend Alan Peacock, who is perhaps better known as designer of the lettering on the Beatles’ “Rubber Soul” LP record album).
Turnbull has pointed out that until a few years ago he and Jennison thought their book represented a definitive solution (“game over”) but an alternative method has gained momentum in practice. Jennison’s talk noted that this alternative may be inefficient, and showed how it can be improved along the lines of his earlier work with Turnbull.
Dr. Hall’s talk, introduced by Russell Barton, dealt with how government agencies, notably the Food and Drug Administration, regulate and guide clinical trials, for example of antiretroviral drugs. In his introduction Professor Barton thanked Turnbull for his patience and compassion, illustrated by Turnbull’s patient work with Barton on structuring his thesis research, “a model for me in advising my own Ph.D. students,” Barton said, and by Turnbull’s compassion in taking him on as a thesis student. At the time Barton was working elsewhere as a full-time consultant on recidivism for the Department of Justice, making him a risky prospect as a doctoral student who had never taken a course from Turnbull and had no idea about the specific theoretical work he might undertake.
Dr. Anderson has developed clinical trial design software with core calculations based on an algorithm by Turnbull and Jennison. His software is now part of the open-source ‘R’ statistical package. Introduced by Turnbull’s ORIE and Statistics colleague Professor David Ruppert, Anderson discussed extending the trial design methodology to account for certain characteristics encountered in cancer trials, such as the occurrence of multiple ways to define the endpoints of the trial, multiple dose groups or combinations of treatments, and multiple biomarkers, including significant genetic sequences (i.e. BRCA1 or BRCA2) that may be present across the spectrum of treated individuals. The result may be a multiplicity of hypotheses, some of which may be rejected at different points in time as data are collected during the trial. To cope with the resulting design complexity, Anderson described his work with coauthors to extend a graphical approach in which hypotheses are represented as nodes, initial significance levels are attached to the nodes, and these levels propagate through edges connecting the nodes as the trial progresses. He illustrated the approach through a case study comparing two treatments for a specific type of cancer.
Professor Louis, who recently completed two years as Associate Director for Research and Methodology at the U.S. Census Bureau, contrasted aspects of design approaches used in epidemiology and biostatistics with those used in surveys such as those that are the bedrock of the US Census. He concluded that “Increased cross-fertilization between the epi/biostat and survey domains will benefit science and policy.” He was introduced by Martin T. Wells, the Charles A. Alexander Professor of Statistical Science in Cornell’s Department of Statistical Science. Louis said that “Bruce is a complete professional and individual. He melds technical and subject-matter skills, transmits our culture and norms to the next generation, is trusted and trustworthy; welcoming and generous. The retirement event mirrored these qualities.”
Turnbull’s wife Martha (Marty) retired in 2006 as Director of Human Resources (HR) at Ithaca College and has had an HR consulting firm since then. ORIE Director David Shmoys presented her with flowers at the conclusion of the event. She has reported that “Bruce has been phasing into retirement and already is spending more time on his favorite leisure activities: tennis, sailing, boating, skiing, and play time with grandkids, Blaise, Seneca and Lochlan.”