Dr Geert Molenberghs is Professor of Biostatistics at Universiteit Hasselt and Katholieke Universiteit Leuven in Belgium. He has co-authored many books, including Missing Data in Clinical Studies with Michael Kenward for Wiley and is a Co-Editor of Wiley StatsRef, our online major reference works resource in statistics. He is the Founding Director of the Center for Statistics. He is also the Director of the Interuniversity Institute for Biostatistics and Statistical Bioinformatics (I-BioStat).
Dr Molenberghs is also Chair of International Relations for the American Statistical Association and a former President of the International Biometric Society. His main research interests have included surrogate markers in clinical trials, and on categorical, longitudinal, and incomplete data
Alison Oliver talks to Dr Molenberghs about his career in biostatistics and clinical trials.
1. You obtained a BS in Mathematics followed by a PhD in statistics both at the University of Antwerp. What was it that first introduced you to statistics as a discipline and what was it that led you to your decision to major in the subject?
It’s a bit of a complicated history. I studied what we would now call Bachelor in Masters but in those days we still had the old terminology. Anyway, I studied mathematics in Antwerp and I actually chose to start the PhD in Algebra. But these were still the days that military service was mandatory, but you could opt for an alternative route, which was civil service, as we called it then. Military service was one year and if you wanted to do civil service, it was 20 months. My PhD advisor knew a professor and biostatistician at the University of Leuven in the Faculty of Medicine, who had applied for somebody doing civil service to help him out with some statistical work. I told my advisor that I planned to do civil service after my PhD, so he put us together. From one thing came another, and we decided jointly not to wait until the end of my PhD but to actually interrupt it, and then I would do statistical intermezzo and go back to pure mathematics.
It turned out to be 16 months of civil service, because it was within a care organization – the academic hospital in Leuven. I really liked the combination of mathematical ideas and the rigour, but also the real data problems and the programming. I then did some follow up administration to formalize my conversion to biostatistics and I rerouted – officially staying at the University of Antwerp, but my day-to-day supervision was at the University of Leuven. So that’s how I got into the field: I would say, by accident!
2. What led you to focus your career on biostatistics and clinical trials?
I liked the combination of mathematics, computer science, data analysis, the real data problems. But the real problems actually came from epidemiology and clinical trials, because the Department of Biostatistics, as it was then called, at the University of Leuven, was working mostly on medical statistics, biostatistics and clinical trials. I did like, on top of that, the particular framework of clinical trials.
During the first two years, I got exposed to applications in other areas, such as social science and psychometrics, because my thesis advisor at the time, Emmanuel Lesaffre, took me to several modelling conferences. I particularly liked that it was not only an area I enjoyed working in but that you also had that sense that you were actually doing something that may be helpful to society, for people, for patients, for populations, making healthier work towards better food supply, better agriculture and forestry, etc.
3. Your main research interests have included surrogate markers in clinical trials, and on categorical, longitudinal, and incomplete data. What are you working on currently?
My initial work was in categorical data, so from clinical trials, you need various multivariate categorical data, but in the repeated measures of longitudinal complex, as you could call it, this mathematically brings a missing data aspect. Very early on, when I was at the University of Hasselt, I met Mark Boyce, and he had a little problem on the evaluation of surrogate markers. But rather than solving his problem, I said, “Wait a minute,” and we started talking about the field as a whole, and then from one thing came another, and we are now more than 20 years later, still working on those surrogate markers with a large group of co-workers.
What has changed, of course, is the data volume that we can handle. Some members of our team would not look at a single surrogate marker but would maybe look at a few million. So there is a big data aspect coming into the picture. So the volumes have changed, the scale has changed, but the substantive areas are in a way still very similar.
4. You have taught at the University of Hasselt & Ku Leuven for over 20 years. What is it about the institutions that you love, what has kept you there?
That’s a good point, because I’ve been at Hasselt now since 1993. I obtained my PhD in early 1993 and then the first few months after that were used to apply for various positions, and I eventually got appointed at Hasselt. However, before fully taking up my position, which didn’t happen until 1st January 1994, I spent half a year at Harvard, in their Biostatistics Department. That, of course, was an eye-opener, because in those days, at Harvard, only in the Biostatistics Department and its affiliated institutions, there were more statisticians than you could find in Belgium and perhaps the Netherlands together! It was on a huge scale to the European eye with their Cancer Clinical Trials Unit and AIDS Clinical Trials Unit.
I had international contacts already then, also through my thesis advisor, with people in the United Kingdom in particular, such as Mike Kenwood, with whom I’ve collaborated my entire career. What did attract me to Hasselt in the first place was the fact that they were running an MSc in Biostatistics and the idea was to supplement the MSc, which is teaching with a research line in biostatistics samples and my assignment was to at least contribute to the development of that research line.
I came from Leuven and kept working with people there, and over the years we reinforced the links that eventually became a strong umbilical cord between Hasselt and Leuven in terms of biostatistics. We exchanged research and students, continually working together. There were many people in one university who took courses in the other. We built courses in both universities until we decided, just over ten years ago, to start an inter-university institute. In order to make that happen in a consolidated fashion, I was partly reappointed to Leuven, so now I’m 50-50 Hasselt-Leuven. I think at any point in time the two universities together have something like 300 masters students in statistics. Even at the teaching level, the two MA programs have a memorandum of collaboration. At one point in time, we ran the only two maths programs that had an accreditation from the Royal Statistical Society. Of course, many UK MScs are now accredited by the RSS, such as several overseas institutions, typically from the former commonwealth, like Hong Kong, etc., but we were proud to say that we were amongst the first to have that, let’s say, stamp of quality. All of these things together sum up the reasons to be where I am and to be happy where I am.
5. You are the Founding Director of the Centre for Statistics, as well as the Director of the Interuniversity Institute for Biostatistics and Statistical Bioinformatics (I-BioStat). Please could you tell us more about how the aims and scope of the Centre and I-Biostat?
We decided it was better to collaborate rather than to compete. We follow a subsidiary way of looking at things that every entity can do separately. We made that more difficult by forcing it to be done jointly, but there are cases, for example where we say, “This is a fairly big consultancy contract, neither of the two entities are able to do it separately. So they will go for it jointly.” We actually have secured several consulting assignments from industry, government, etc. that we could not have done individually.
But we also noticed that ever since we began to work together that both MA programs have increased in size. I think if it’s clear that we collaborate, then we can be even more of a point of attraction than forcing people to make a difficult decision as to which institute to go. It’s a very natural construct.
6. How do you feel the teaching of statistics has evolved and adapted to meet the changing needs of students since you began teaching?
Pretty soon after I began my career, the human genome revolution came into being, so there was statistical genetics and bioinformatics; now we talk about data science, big data, etc. These evolutions were there all the time, and they forced statisticians to think very carefully.
There is a delicate balance between two extreme views. One view is there’s an entirely new field called data science that’s emerging and may replace statistics. The other extreme view is: it’s business as usual. And it’s neither of the two, of course. This has happened before, in the middle of the previous century when all of a sudden, epidemiology and observational studies came into being and the classical experimental paradigm had to be adapted to which Fisher, Biersten and all the contemporaries had contributed. It’s not a new field, but it’s a natural evolution of a field. So I think from a more distant perspective, what you see happen is that there are methodological and technological evolutions in teaching facilities. When I started teaching, you were very advanced if you were using a slide projector, because still then a number of professors would use nothing but the chalkboard. If you would come in with overhead transparencies these days, your students would look at you as if you were a dinosaur. Now you can use laptops and iPads—I often use my iPad in my teaching. And in particular, there are now more distance learning opportunities, the use of web lectures, etc. I think all these innovations are very beneficial for statistics, especially the flipped classroom where students watch web lectures, as statistics is after all a technical field. I always say if I gave a theorem and a proof, which I don’t do all the time, but it does happen, you can do it at two paces: either too fast or too slow. But if you supported blended learning tools to slideshow lectures, students can literally go through the material at their own pace. For some where it’s too slow they can fast-forward a little bit, and others where it’s too fast, they can push the pause button or slow it down or rewind, etc. Both methodological as well as purely pedagogical innovations have pretty much changed the way statistics is taught.
7. You were President of the International Biometric Society from 2004-2005. What are your memories when you look back on your Presidency?
It actually happened somewhat coincidentally. I had an involvement in the Biometric Society in a variety of ways. For example, I served on the board of the Belgium region, which is in Belgium called The Quetelet Society, because Auguste Quetelet was a Belgian. And by the way, parenthetically, he was the first foreign member of the Royal Statistical Society. There is a statue of Quetelet for that reason on Errol Street where the RSS headquarters are. I was the treasurer for quite some time, and then at one point I was visiting France, and I was asked by the then President of the IBS, Brian Morgan, to become the next Editor of the Biometric Bulletin, which I agreed to. Being Editor was a fine experience because you got to know almost first-hand all the activities happening within this very kaleidoscopic society with national and sometimes supranational regions. It’s not like one monolithic society; there are 50 more or less regions and groups within the society in our main database, the North America region being the largest one which is roughly 40% of the membership. I also obtained an appreciation and an understanding for the diversity in type of biometric and biostatistics activities that are done in different parts of the world. I learned for example, 20 years ago the UK was still very much focusing on agricultural biostatistics, other countries on medical statistics.
After that period, I served as General Secretary, and the President of the time was the late Norm Breslow, and his successor, incoming Vice President, was Rob Kempton, from Biostatistics and Mathematics Scotland. Unfortunately, Rob Kempton died during a sad first year, and that was a terrible shock because he was in very good shape, had just completed an excellent visitation cycle of the institution, and then he suddenly passed away. Elections had to be organized, and from 2003 onwards I knew I would be the president for 2004-2005.
This was at a different level of course, but a continuation of my experience with the society as I had had as the editor of the Biometric Bulletin, meaning you get to appreciate no longer corresponding on written pieces but from seeing people face-to-face. I got to appreciate the wide diversity and variety of the international society, and I think something like this helps you at least become more able to understand why somebody from Latin America looks at things this way, and somebody from the US might look at it differently. It’s a very rich genetic makeup. Needless to say, of course, if you’re in the driving seat so to speak, there is also administration and politicking, but the major dominant positive experience for me was the international flavour of the society, which comes together once every other year in the International Biometrics Conference. When I was President, that took place in Cairns, which is a wonderful part of the world, being located next to the Great Barrier Reef and the rainforest in Australia. I have very fond memories of my time there.
If not every type of limitation and barrier, wall we could say, is put in place, that goes against the willingness to continue then to collaborate internationally, and I think at a more profound level, no matter what happens politically, it’s a good idea that scientists try to collaborate with whoever else wants to collaborate in our field. Whatever happens politically, if you are collaborating with people in country XYZ, and the goal is for everybody to develop beneficial methods and materials, such as better medication, improving food supply, etc. that statistics can contribute to, then everybody benefits and to the extent possible we try to make that happen.
8. You are also Chair of International Relations for the American Statistical Association. Please could you tell us more about your role, which also included a recent issue of a statement in response to President Trump’s Executive Order on visas and migration.
There are two national societies in statistics, which are at the same time international societies – the Royal Statistical Society and the American Statistical Association. Over the years, I got to know and work with the RSS and ASA quite a bit, and in different categories for different things, such as serving on their research committees and I was Editor of Series C of the RSS Journals for a while.
From one thing came the other, and they asked me, because I’m a frequent attendee to the joint statistical meetings, to become the chair of the Committee on International Relations in Statistics. And virtually all—not all, but virtually all—members are either based at international locations. They may well be American based in international locations, such as people on the committee who work in Africa and have been dispatched by, for example, the CDC the Centers for Disease Control and Prevention in Atlanta, or are people working in the US but are from Columbian descent, for example.
The major activity from the committee is to select every year a so-called educational ambassador. What does the educational ambassador do? It’s somebody who is well connected in their local community; for example, the first one came from Argentina. It was somebody with good connections to the local framework, the local network, the local organism, so to speak, of statistical nature. That person then attends a joint statistical meeting section, selects a number of short courses, and follows those with a lot of attention and goes back and basically imparts the knowledge on the country. That’s essentially the idea. Now over the years, we have refined it; we are also using electronics like web lectures. We tried to establish links between the short course instructors at JSM and the educational ambassadors so for example, the person who is teaching a short course at JSM can receive further input and support from the educational ambassadors.
Of course, you are really looking in that committee at the international segments of the society, and not only that, of the association. I should say, in case of the ASA, you are also looking at where activities of the ASA can be extended, expanded, and you’re looking for opportunities where the ASA can be at service and be helpful in building the local statistical network. If not every type of limitation and barrier, wall we could say, is put in place, that goes against the willingness to continue then to collaborate internationally, and I think at a more profound level, no matter what happens politically, it’s a good idea that scientists try to collaborate with whoever else wants to collaborate in our field. Whatever happens politically, if you are collaborating with people in country XYZ, and the goal is for everybody to develop beneficial methods and materials, such as better medication, improving food supply, etc. that statistics can contribute to, then everybody benefits and to the extent possible we try to make that happen.
9. You have authored many publications. Is there an article or book that you are most proud of?
It’s difficult of course, because there are many that I really like and I look back on with certain pleasure.
I think in particular I would like to mention the first book that I did with Geert Verbeke, the one on Mixed Models in the Springer series. One of which I have very fond memories also is Missing Data with Mike Kenward in 2007, which was published by Wiley. I think the advantage of a book is you can take a few more words than you can in an article or reference paper, to bring your ideas together to expand your ideas. So it’s like a second and a third, and if you write more, a fourth and a fifth PhD thesis, but you hope that every time you get a little better and if not at least you gain work. Consolidating a good amount of your own research and that of others, of course, in a book that is also used as a textbook, is a pleasant experience. I think it’s really the simplicity that comes with it which makes it so nice.
10. What has been the best book on statistics that you have ever read?
That also is difficult to say, because there are several. I would certainly include Categorical Data Analysis by Alan Agresti. For me, it’s the paradigm of a textbook. It’s a well selling book. Alan has revised it over the years a couple of times, but the essence, the foundation, the core is still there and it still stands the test of time.
Another book I should like to mention is Allen Welsh’s Aspects of Statistical Inference. As a statistician, we of course know that inference is the core foundation of our field. If inference wasn’t there, there would be nothing. It would be a house of cards. There are, of course, very solid and thorough textbooks out there but they are not necessarily fun to read. Allen Welsh’s book is one exception to that. He understands the art, not only of going a little further than what many other textbooks do and provides more detail, such as historical connotations. For example, if he talks about likelihood inference, even though he will describe what it is, he’ll also go into how it came into being and the nature of protagonists and antagonists, faults with each other and follow the dialogue throughout, and he would discuss even defunct methods of inference that most people wouldn’t mention because it’s barely used and for good reason. If you read that book, you really feel like you were there when the developments took place—it’s that well written.
11. Who are the people that have been influential in your career?
Emmanuel Lesaffre was my day-to-day practical thesis advisor, so of course he has been very important for me. Mike Kenward, with whom I’ve worked ever since the beginning of my PhD, because Emmanuel brought to my research his agenda—categorical data, multivariate categorical data, etc.—but Mike added the flavour of longitudinal data and mixing data. As we were PhD students together ever since the beginning, I did a lot of work jointly with Geert Verbeke. We remain close friends, and I collaborate with him all the time and we still work together, and even co-direct the Institute together with a few other people.
Herman Callaert was the founding chairman of the Masters in Statistics in Hasselt University, and he was the one together with his colleagues Noël Veraverbeke and Paul Jansen who hired me. Herman was the one who really insisted on me going to Harvard, before I really took up my position. To this day, even though he retired ten years ago, he is a mentor to me.
Louise Ryan, the incoming president of the Biometric Society and the editor-in-chief of Statistics in Medicine, whom I knew before meeting her again at Harvard as she taught a course in Hasselt a guest lecturer. I also worked a lot with Stuart Lipsitz, and Garrett Fitzmaurice, publishing quite a few papers together; Garrett was involved in the two handbooks that we did for Chapman & Hall on the nature of data and on missing data.
Stephen Lagakos was also important for me because he was a good friend and he was my host at Harvard. He was one of those people who were very busy but when you had the opportunity to talk to him, he gave you the impression that he had all the time in the world for you. If you had an appointment with him at 4pm, you would be waiting until 5.30pm! But you knew he would make the time for you that you needed and deserved. He was a mentor for that half year. I had three or four conversations with him. It doesn’t sound like a lot, but each one of them was memorable, and still to this day I practice the lessons that he taught me. Unfortunately, he passed away in a tragic car accident in 2009. I will never forget him.
12. If you had not got involved in the field of statistics, what do you think you would have done? (Is there another field that you could have seen yourself making an impact on?)
It’s a little bit of a counter-factual question in causal inference. So the question is what would you do in a parallel universe? An obvious answer would that I would have been a pure mathematician. But up to this day, I’m not really sure whether I would have stayed in that field.
When I was a teenager, I thought of becoming a civil engineer. When I was a child, like six or seven years old, I dreamt of becoming a cook, and that lasted for a few years, and then that was gone. And in a kind of a parallel life, knowing it would never happen, I saw myself being a tram driver, and the reason actually is because public transport, rail transport in particular, has always been a great hobby of mine. It also sparked my interest in history, like how did the underground in London come to be? How did the metro in Paris come into being? How did the subway in New York arise? At one point, you have to think about what you should do when you retire? Well something combining all of that, let’s say fifty percent being a cook and fifty percent percent being on the historic transportation side would be something that I could see myself doing.