key: cord-0712662-z2dhkz6y authors: Harel, Ofer; Zigler, Corwin title: A conversation with Thomas (Tom) R. Belin- 2020 HPSS long-term excellence award winner date: 2020-07-01 journal: Health Serv Outcomes Res Methodol DOI: 10.1007/s10742-020-00212-w sha: 202ca66103836cd9d28e0a411a654df6035fe264 doc_id: 712662 cord_uid: z2dhkz6y At the 2020 International Conference on Health Policy Statistics held in San Diego, Thomas (Tom) R. Belin was awarded the Long-Term Excellence Award from the Health Policy Statistics Section of the American Statistical Association. Dr. Belin was exceptionally and uniquely qualified for this award. Highlights include his innovative statistical applications for health care research and his substantial contributions to the statistics and health policy communities through mentoring and service. In this interview, we asked Tom to share stories about his upbringing, schooling, and career phases to gain insights into his numerous achievements. The HPSS Long-Term Excellence Award recognizes significant contributions to the profession through mentoring and service. Candidates must be 15 or more years from their terminal degree on January 1, 2019. Tom met the eligibility criteria for the 2020 award as he earned his Ph.D. from the Harvard Department of Statistics in 1991. Having spent most of his career at UCLA, he serves as a Professor in the UCLA Department of Biostatistics and UCLA Department of Psychiatry and Biobehavioral Sciences. Along the way Dr. Belin served for a year as Acting Chair of the Department of Biostatistics and currently serves as the departmental Vice Chair. In research, Dr. Belin is internationally known for contributions to address practical problems in incomplete data. His methodological developments have often drawn on models incorporating latent structures and Bayesian estimation strategies, and in addition to health-related applications he has made lasting contributions to the literature on record linkage. He has secured grant funding from a wide range of government agencies and private foundations, and his 165 peer-reviewed publications, which include numerous contributions to top-tier journals, have garnered more than 15,000 citations. In leadership and service, after being a member of several organizing committees for the flagship conference of HPSS, he co-chaired (with Arlene Ash) the 2005 installment of the ICHPS series, and he has subsequently been a member of ICHPS advisory boards. Through his many other service contributions in academia and professional societies, he is known as a bridge builder across institutions, communities, and individuals. Within the ASA, he has held leadership positions in the Survey Research Methods Section and the Biometrics Section, and he has served on both the ASA Census Advisory Committee and the ASA Committee on Professional Ethics. His professional contributions have been recognized by a diverse array of organizations (including the Oral and Maxillofacial Surgery Foundation, the American Academy of Child and Adolescent Psychiatry, the International Center of Mental Health Policy and Economics, and the UCLA Public Health Student Association), and recognition for his role in developing the Community Partners in Care project on dissemination of evidence-based strategies for depression in under-resourced communities included the Team Science Award from the Association for Clinical and Translational Science and American Federation for Medical Research, the Annual Award from the Community-Campus Partnership for Health, and the UCLA Community Program of the Year-Landmark Award (Belin et al. 2018; Shetty et al. 2008; Wells et al. 2013; Zima et al. 2010) . Other recognition for his career contributions include being elected a Fellow of the ASA in 2004, receiving the Gertrude M. Cox Award from the Washington Statistical Society in 2005, and being the Lowell Reed Lecturer at the 2018 American Public Health Association annual meeting. The remainder of this paper draws on a recent conversation that Ofer Harel and Cory Zigler had with Tom that included plenty of lighthearted anecdotes while probing into the reasons for his success and development into the wonderful academic colleague we see today. Questions are in Roman font while Tom's responses are indicated by TB and are in italics. Can you summarize your upbringing? Where did you live as a child? What did your parents do and how did they influence you? TB I had an idyllic childhood growing up in Des Moines, Iowa as the fourth of five kids of parents who were pillars of the community, benefiting as well from strong influences from other neighborhood families. There was always a lot of action in our home with so many kids in motion, and one reflection of the environment that you wouldn't see in today's society was a sign in our neighbor's backyard that served as the baseball or football field depending on the season where Rule Number 1 was "Everyone is welcome to play. " (Belin 1973 (Belin , 1988 Tell us a bit about how Stanford became the destination for your undergraduate studies and about your choice of major. TB Both my older brother Jim (who majored in Political Science, graduating in 1980) and my sister Joy (who majored in History, graduating in 1983 , and stayed for a masters degree in Education) had gone to Stanford, and I had enjoyed visiting them there, so once I got in it was an easy choice. That said, I had always viewed college as the stepping stone to law school, with my father as a role model, and originally, I was planning on being a History major. In fact, I remember finishing my freshman year thinking I would never take a math test again. But that summer, when I looked more closely at the History major requirements, there was a requirement beyond the history coursework that could be satisfied with either more of a command of a foreign language than I had or with a year's worth of statistics. Given what seemed to me to be a clear choice, I browsed through the catalog of possible majors to see whether there was an option to combine the math I had already taken with the statistics I was planning to take, and I came across the interdepartmental Mathematical and Computational Sciences major, which combined math, statistics, computer science, and operations research and which seemed just right. In meeting with the head of the program to be assigned an advisor, I was asked what I wanted to do going forward, and I remember saying that I didn't know but that I had thought about going to graduate school and had thought about teaching. The head of the program, Brad Efron, responded by saying something close to, "I think I'll keep you as one of my own advisees to keep an eye on you." I did not know who Brad Efron was at the time, but that was a life-changing, butterfly-wing-phenomenon moment for me. I would meet with him every quarter, and I also was mentored by Herb Solomon, who allowed me to take an independent-study course on statistics and the law with him after I had him as a professor in one of the core courses. Were you thinking of an academic career at this point? If not, what else were you thinking about? TB When I declared the Mathematical and Computational Sciences major, I still thought that law school was an option. But my older brother, who graduated from law school at the end of my freshman year of college, didn't seem to be enjoying his work as a lawyer that much (he eventually went to business school and pursued portfolio-management work). I also liked working as a math tutor, and I found myself thinking that a direction aligned with the scientific coursework I was taking would be a better fit for me than a direction with less precision and more ambiguity. I could tell that there was an adjustment at first for my father when my plans solidified around graduate study in statistics, but as events unfolded he clearly enjoyed the new horizons I was pursuing. You were part of an interesting cohort in graduate school. Can you describe what that was like and whether, all these years later, you still trace your particular professional path back to that cast of characters? TB It was an amazing experience to be at Harvard during those years. The group that entered when I did included Andrew Gelman, Xiao-Li Meng, and (Gelman et al. 2013) ]. Andrew, Xiao-Li, and I were all working with Don Rubin, but on different things. Andrew's Ph.D. thesis was on combining PETscan reconstructions in experiments (which vaulted him into the realm of MCMC computing), Xiao-Li was developing the SEM and ECM algorithms as extensions of EM, and I was working on calibrating false matches in computer matching related to census undercount estimation. It also happened that in the fall of 1986, when we started, the Red Sox were in the World Series, playing the N.Y. Mets in the series famous for their Game 6 collapse when the ball went through Bill Buckner's legs. Andrew and I went down to Kenmore Square to watch Game 3, the first in the series played in Boston, in one of the Fenway Park bars; the Mets hit a leadoff home run on their way to 4 runs in the top of the first inning and a wipeout victory, and more memorable than the game was the short course in human diversity we received from the people crammed into the same booth, such as from the woman who mixed tomato juice into her beer. Later I took Xiao-Li to his first baseball game, where I remember spending almost the entire time explaining the rules of the game. And then in the ninth inning, a rule was invoked that I knew existed but that I had never seen invoked before and have never seen invoked since, involving the one situation when someone can be put back in a game after having been taken out (namely, when a catcher is replaced and then the replacement gets injured, the original catcher can come back). I remember trying to explain to Xiao-Li not only the substance of the rule but also how rare it was for it to be invoked. To put the timing of our graduate studies in perspective, the first edition of the Little and Rubin "Statistical Analysis with Missing Data" text came out in 1987 (Little and Rubin 2019) , as did Rubin's "Multiple Imputation for Nonresponse in Surveys" (Rubin 2004) . To put the experience of being a classmate of Andrew and Xiao-Li in further perspective, I'm reminded of an anecdote growing out of our Ph.D. qualifying exam. Drawing on Don having taught a class out of his newly published multiple-imputation text, I remember recognizing one of the problems as being related to the Bayesian Bootstrap, and I thought I had just nailed it. But when I had an opportunity to review my graded exam, the only marking on the page when I got it back was "9.5/10". Unable to resist asking the question, I went into Don's office soon after and conveyed that I was curious why I got 9.5/10. Without skipping a beat, Don said, "Oh, that's because at one point in Xiao-Li's answer, he wrote that such-and-such was obvious from a Bayesian perspective." And I thought, "I see-it wasn't that I did anything wrong, it was just that his answer was better!" I thought that was a good lesson, and I've shared the anecdote with many students over the years in situations where it is possible to do better. Tell us about some influential work you did while in graduate school. TB My thesis work on record linkage that was summarized in Belin and Rubin (1995a) estimating error rates by drawing on previous matching experience to fit mixture models, didn't get replicated a lot, but it did get recognized as an interesting idea that broke new ground, so articles in the record linkage literature frequently cite that work. The censusrelated project fitting a hierarchical logistic model to resolved post-enumeration-survey cases as a way to predict enumeration status for unresolved cases led to a 1993 JASA discussion paper (Belin et al. 1993 where the schizophrenic reaction times were posited to come from a two-component mixture distribution to account for their greater variability. The article summarizing that work (Belin and Rubin 1995b ) includes a lot of technical detail on applicable EM, ECM, SEM, and SECM algorithms, which might be finessed or bypassed in the MCMC computing era in favor of an iterative simulation algorithm. But a part of the paper that still holds up well is the illustration of how posterior predictive checks could be used to help refine the mixture models to accommodate salient features of the data, and I still present that illustration to my own advanced students. Kruskal what data he would need to see in order to favor adjustment, and I asked the other panelists what data they would need to see in order to oppose adjustment. The assembled audience immediately burst into a roaring applause-the discomfort I had felt about judgments being made without any reference to relevant data, which were still in the process of being collected, was clearly shared by others. That was a great moment for me, as I felt that I had taken a stand for science and that I was entering a profession with a great many kindred spirits. (I believe in 1989) , called me early in 1991 to tell me that there was an opening for a post-doc at UCLA. Apparently, there had been a conversation in the background between Nat and Bob Elashoff, who eventually hired me, that set events in motion. Arlene on a "hi-how-are-you" basis for many years, as she would often join for seminars at Harvard from her position across the river at Boston University, but I didn't know her very well prior to co-chairing the 2005 meeting. I was pleased to say "yes", as it seemed like a great opportunity, and in the process, she and I hit it off and have been close ever since. What is the favorite paper you have been involved in? Or the one that you are the most proud of? TB That is hard. I'm going to not name a favorite one, but a set of papers that are meaningful. One was this paper on the Census Bureau work that included Joe Schafer, Don Rubin, Alan Zaslavsky, Gregg Diffendal, and Steve Mack (Belin et al. 1993) . I felt like that one helped launch my work into more sophisticated methods. The two others I mentioned from my graduate study, both with Don Rubin (Belin and Rubin (1995a) , JASA paper on record linkage and Belin and Rubin (1995b) , Statistics in Medicine paper on mixture models for schizophrenic reaction times), really helped me understand incomplete-data methods and posterior predictive checks, and I've referred to them a lot in my teaching. One other paper that stands out from earlier in my career is on DNA identification. I had been sitting in on Ken Lange's course on statistical genetics, and he was giving a presentation on forensic DNA applications, where he had presented a method for relaxing reliance on gene-frequency databases in calculating match probabilities, and as a headstrong post-doc I had said, "No, no, no, that's all wrong- the defense lawyers will pick on you for any conditional independence assumption-you should use a distance metric on the distance between the bands on these assays and finesse the independence assumption." I think he was a bit startled by my soapbox declaration, but Ken's thoughtful response was, "Well, write it up!" It took a while, but I worked with David Gjertson, who had access to genetic paternity data, and my student Ming-yi Hu, and we did write it up. Along the way, while that paper was in pre-print form, the O.J. Simpson trial was going on, and Bob Elashoff encouraged me to send the paper to Judge Lance Ito of the court, which I did. In my files somewhere, I have a reply letter from Judge Ito saying he'd give the paper to both the prosecution and the defense. We first submitted the paper to the American Journal of Human Genetics, and although the comments were mostly vaguely positive, the paper was rejected on the grounds that it didn't make any use of genetics theory. But that was exactly the point! So we sent pretty much the same paper to JASA Theory and Methods (Belin et al. 1997) , and they loved it because it didn't make any use of genetics theory. I should also mention, in addition to your dissertation papers Belin 2011, 2012) , Cory, that one other paper I'm especially proud of is the paper we worked on with Heidi Fischer developing density-variation/compactness scores for evaluating redistricting plans ). It's far from among my most cited papers at this point, but I still think of the idea as having merit and as having great potential to be influential, and I see myself doing something with it during the next round of redistricting that's coming up. It seems that you know "Everyone"-how has that helped/hindered your career and/or dayby-day life? I had started out as one of the Co-Recording Secretaries, taking minutes at half of the meetings. I don't know if it was because my minutes read like hours, but a couple years later I got asked to be the President. I didn't think I'd be able to do it, but said I would if they could find a Co-President. They did, although about a year into a two-year term, the Co-President dropped out. Still, those were great experiences. It was demanding at times, but I am proud of that involvement, and I feel that plugging in the way I did, which included increasing the PTA budget by a six-figure amount to support a reading specialist and to cover for other funding cuts, had a meaningful impact on the school and on the community. In my down time you'll also find me reading, listening to music, going to farmers markets, on a good day going for a short run or a long walk, and thinking about ways to make the world a better place. Dr. Tom Belin's accomplishments are too many to count, the number of students, colleagues and mentees that will be happy to praise him is longer than we can mention. We are confident Tom will continue producing influential research via papers and grants. There are many more students and mentees who will be lucky to be mentored by Tom. If you attend statistical conferences, there is a good chance you will cross his path and you can hear his roaring laugh and his beaming personality. You Are The Jury. Quadrangle/The New York Times Book Final Disclosure: The Full Truth About the Assassination of President Kennedy. Charles Scribner's Sons A method for calibrating false-match rates in record linkage The analysis of repeated-measures data on schizophrenic reaction times using mixture models Hierarchical Logistic regression models for imputation of unresolved enumeration status in undercount estimation Summarizing DNA evidence when relatives are possible suspects Using a density-variation/compactness measure to evaluate redistricting plans for partisan bias and electoral responsiveness Maintaining internal validity in community partnered participatory research: experience from the community partners in care study Strange But True Baseball Stories. Random House Statistical Analysis with Missing Data Multiple Imputation for Nonresponse in Surveys Analysis of Incomplete Multivariate Data Do the benefits of rigid internal fixation of mandible fractures justify the added costs? Results from a Randomized controlled trial Community-partnered cluster-randomized comparative effectiveness trial of community engagement and planning or resources for services to address depression disparities The potential for bias in principal causal effect estimation when treatment received depends on a key covariate A Bayesian approach to improved estimation of causal effect predictiveness for a principal surrogate endpoint Quality of care for childhood attention-deficit/hyperactivity disorder in a managed care medicaid program Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.