Adrian Bowman is a Professor of Statistics and Head of the School of Mathematics & Statistics in the University of Glasgow. His research interests include environmental modelling and the analysis of anatomical shapes. Professor Bowman also has a longstanding interest in educational technology, particularly in the role of graphics in aiding the understanding of statistical ideas. Over the years, he has had many roles within the Royal Statistical Society, including joint-editor of Applied Statistics, Honorary Secretary of the Research Section, member of Council, chair of the Careers Promotion Committee, Honorary Officer for Membership and Guy Lecturer.
Alison Oliver talks to Professor Bowman about his career in statistics so far.
1. Please could you tell us more about your educational background and what was it that inspired you to pursue a career in statistics?
I grew up in Scotland and went to the University of Glasgow to do a mathematics degree. I was considering what to do next and didn’t feel that doing more mathematics per se was necessarily going to be for me, but statistics was a natural direction to go in, as I hadn’t done much at the undergraduate level. It seemed a good route forward both in terms of employment and relevance to the world. So, I did a 9-month course in Cambridge – the Diploma course as it was called then – and after that decided to do a PhD. I ended up coming back to Glasgow for that, to work with Mike Titterington – a very well-known name in statistics. That led to a career as an academic – first as a lecturer at the University of Manchester for 5 years and then back in Glasgow.
2. Your research interests include development of inferential tools for flexible models and the application of these methods to environmental data, as well as statistical shape analysis. What are you working on currently?
That’s quite a good summary of the two current themes. If we take the environmental or spatiotemporal modeling one first, that has developed out of a long-standing interest in looking at what I like to call “flexible models” – models which relax the usual assumptions. Of course, that has been a very strong theme in statistics for many decades, and I have found it an interesting and rewarding area to work in. Most recently, it has expressed itself in problems which have many variables and often have a spatiotemporal component – and there are many applications of that in the environmental arena.
I can think of two projects at the moment that encapsulate this. One is with the Scottish Environment Protection Agency, who hold data on water quality in the River Clyde stretching back decades. There are really important questions about what’s going on in the river system – a very important part of the human and natural environment — because, not to put too fine a point on it, its where discharged sewage goes. This is fully treated and there is no human health risk here, but because we are dealing with a large population any effect on the quality of the water has to be very carefully monitored. So, we have been working with SEPA to analyse the historical data on this. That involves not only thinking hard about a suitable statistical model, but also to think about the questions that are being asked and making sure the model is consistent with what is already understood about the river system. This becomes quite sophisticated, because a river system is a complex system – it’s affected by the time of year, the short-term condition of the river, by where you are on the river, and of course by any long-term trend, which actually is one of the most important things to be able to identify. So that’s really fascinating, and another good example of quite a sophisticated statistical model being attached to a really interesting application. I think that encapsulates why statistics is so vitally important.
The other project in that area at the moment is with Shell, a multi-national company. For some years we’ve been working with them on a tool to monitor groundwater pollution, which is very important because you don’t want any industrial activity to have an impact on groundwater which may then seep out and affect the surrounding area. So that’s been interesting, too. The tool (GWSDAT) that’s been developed is already quite widely used, and we have a project to move it to a new level and increase its use further. And that again is one of the most satisfying things about statistics. There are opportunities to work on methods, models, theory, and there are opportunities to work on applications, but when it gets to the stage that people are actually using it for real, that’s very rewarding.
3. Your lecture at RSS 2017 was entitled ‘Surfaces, shapes and anatomy’ where you talked about statistical models being applied to the human face. Please could you tell us more about this topic? What is the one thing that you would like your audience to take away from your lecture?
In the end, the message I would like people to take away is a very general one. I said at the beginning of the talk that, for me, working in this kind of area has underlined the fact that statistics is a really rewarding subject, because you get involved not only in the theory, methodology and the modeling side of things, but you also get involved in the interdisciplinary aspects with other scientists and other problems which are really fascinating. Statistics is a really important subject – we all know this. I wanted in the keynote talk to celebrate that fact a little – that statistics is a really important part of almost anybody making a good decision or improving their understanding. The whole idea of quantitatively modelling something, and weighing the evidence for what’s going on, is such a powerful one.
4. What do you think have been the most important recent developments in the field and which may influence your teaching in future years?
Statistics is a subject that really does evolve quite rapidly, and most of my teaching has been done in the University of Glasgow where a Statistics Department was first set up in 1966 and there is a long tradition, unusually, of a full undergraduate degree in statistics. So the kind of teaching I’ve been able to do there has been a little bit beyond the normal provision in a UK mathematical sciences degree. Glasgow has always had a very strong emphasis on combining the theory and methodology with the applications and that is expressed through a substantial proportion of practical laboratory work within the degree courses. That’s true from the elementary introductory courses right through to those who graduate with a degree in statistics. As the decades have gone on, that has increased in importance as the computational tools and the computing environments that are available have improved so enormously.
On a more recent trend, the label for this RSS Conference has included data science, which I think is a very good thing, as statistics is most definitely part of that. One of the trends at the moment is the convergence, to a degree, of what some parts of computing science are seeking to do, and what statistics has always sought to do in terms of analyzing data quantitatively, now potentially with very large datasets. So, it’s going to be interesting to see what that convergence produces. In Glasgow, we have just launched into postgraduate programmes at the taught master’s level which bear the badge Data Analytics, expressing the fact that the course is intended to give a good firm foundation in statistics but also to include elements of computing science that our students on a full statistics program would not normally see. That also seems to be an area that externally is driving a lot of employment.
5. Your research has been published in many journals, books and conference proceedings: is there a particular article or book that you are most proud of?
Everyone has affection for particular papers and publications for all sorts of reasons. Naturally the first few papers you ever publish are high on the list. My very first paper, when I was working as a PhD student with Mike Titterington, was published in Biometrika and there is a lot of satisfaction in thinking back on that one. Another landmark for me was writing an OUP book with my colleague Adelchi Azzalini from University of Padua. That was a very good experience because it brought together quite a lot of the research we had been doing over the previous years and so it was a very satisfying thing to do. The book was quite short, but nonetheless there was a feeling of having produced something more substantial than another paper, with an influence that was correspondingly greater.
Also, by the nature of the kind of statistical modeling I’ve been involved in, I have often published not only in the statistics literature but also in many other journals. For example, a few years ago I had a paper in Science, which is a very highly prized journal in many parts of the physical sciences, as part of a larger group led by a physicist. That one I have particular affection for because my son, who is a physicist, was also part of the author list, so there’s a personal connection. But I’ve also published in places such as the Polar Record, which is a rather unusual place for a statistician and expresses again the fact that we can get involved in the most astonishingly wide range of scientific activity.
One of the trends at the moment is the convergence, to a degree, of what some parts of computing science are seeking to do, and what statistics has always sought to do in terms of analyzing data quantitatively, now potentially with very large datasets. So, it’s going to be interesting to see what that convergence produces. In Glasgow, we have just launched into postgraduate programmes at the taught master’s level which bear the badge Data Analytics, expressing the fact that the course is intended to give a good firm foundation in statistics but also to include elements of computing science that our students on a full statistics program would not normally see. That also seems to be an area that externally is driving a lot of employment.
6. What book(s) have been particularly influential to you?
When I was an undergraduate, I had the privilege of being taught by David Silvey who was the first ever Professor of Statistics at University of Glasgow, and he wrote Statistical Inference, which is quite a short book but very clearly and elegantly written. A number of people have commented to me over the years just how much they like that book, and I understand why. I think it expresses very clearly some of the kind of fundamental ideas of what inference is about.
At a later stage, Generalized Additive Models by Hastie and Tibshirani had a big influence, because that was at a time when I’d been working on flexible regression in simple cases. Looking back now, I would describe them that way, but they didn’t seem simple at the time! The book on additive models synthesized the whole area and made it accessible to a wide audience.
Dryden and Mardia’s Statistical Shape Analysis has been another key book for me because it was a landmark, again in terms of synthesizing work on statistical shape in a way that `raised the game’. It has become a standard reference. All of my postgraduate students who work in shape analysis are given that book to read when they start.
7. What would you recommend to young people who want to start a career in statistics?
Well, I think the first thing I’d say is: yes, go for it – you’ve made a very good choice. It’s certainly a career that I’ve found really fulfilling and enjoyable. It’s evident again from vitality of this conference of 600 people, many of whom are at an early stage in their career. So this is a growth area – the need to model data and understand what’s going on is not going to go away. People who have the skills to do it are really going to have plenty of career options.
In terms of further career advice, I think nobody should offer anything too general here, because it’s always going to depend on the individual. But perhaps the general advice might be to go where your heart is in terms of what you feel is your strength or indeed your calling, if I can use that terminology – where you feel that you’re really interested. That’s the area that you’re going to be most effective in because you’re attracted to it.
The other piece of advice is that you have to learn to balance the competing demands on time in any job. I sometimes reflect back on what the job involved when I started out in my first university post and compare that with what young people are expected to do now. Perhaps young people now are better prepared and smarter than we were but the expectation on what should be achieved quickly is often very high. So good advice would be to give most attention to the essentials of the job: in academic life that boils down to doing research as well as you can, and doing teaching to the best of your ability. Of course, we all have to contribute to making sure things run smoothly too. In general, I would also say it’s very healthy to `cast your bread on the waters’ a little, in the sense of having a go at some things which are not necessarily central to your current agenda. Leave a little bit of time to explore other areas, because sometimes those are the directions that produce really interesting things. But it’s very difficult to balance the short term and the long term when expectations are so high in working life these days.
The other thing is to look for a good mentor. Often nowadays employers will have schemes that support people in the young stages of their career. Without this, it can be quite difficult if you don’t feel you have someone to bounce ideas off or discuss concerns and strategies with. So whatever sphere or occupation you’re in, seek out someone that can help you with that. And of course, join up with your peers. That’s why the RSS is so good, because for young statisticians there are all sorts of activities which links people together at the early career stage. That gives excellent and very healthy support.
8. Who are the people who have been influential in your career?
There are a lot of people who have been influential for me. Going back to the early stages, I’ve already mentioned David Silvey and Mike Titterington—they are both very important figures who taught me. They were both clearly people of very considerable standing in the subject, and over the years I feel they gave me very good advice and support.
For anyone, the first job is always going to be a very influential time of life. Fredos Papangelou was the senior figure at the University of Manchester when I was appointed there and he was very supportive. I greatly appreciated the guidance that he gave at that time.
In general, I suppose advice is really often transmitted by example. It’s not necessarily the things people tell you but it’s the way people behave that’s most influential. It’s usually people who are slightly ahead of you in career stage that you most naturally look to, and there’s a whole raft of very high-profile people who were certainly more prominent than I was, Peter Diggle and Peter Green are both very good examples. I’m sure there are many others I could mention that I really admired in terms of their research work but also, again, it was the interest that they took in other people and encouragement that they gave which was highly influential.
And then of course there’s always a large raft of collaborators. For example, I spent many years working with Adelchi Azzalini in Italy. Those kinds of connections are also big influences because close working relationships develop and you learn a great deal from that.
Copyright: Image appears courtesy of Professor Bowman