Statistician or statistical scientist? an interview with RSS president Peter Diggle

Written by Web News Editor on . Posted in Features-OLD

For someone who has recently been made president of the Royal Statistical Society, it’s perhaps surprising that Peter Diggle doesn’t see himself as simply a ‘statistician’. ‘I see myself as a statistical scientist rather than a statistical mathematician,’ he says. Almost everything I do, I couldn't do on my own, and I like to think that the people I do it with couldn't do it as well if they didn't have me. The statistics are integral to the science rather than a separate discipline.’

This attitude towards the discipline is evident in almost every aspect of Peter’s work. He divides his time between Lancaster University, where he is distinguished university professor of statistics in the faculty of health and medicine, and at the University of Liverpool, where he works in the department of epidemiology and population health.

The two roles, he says, fit together nicely. ‘My operation in Lancaster is very much a statistics research group within the medical school,’ he explains. ‘Liverpool has a much bigger medical school, plus a vet and tropical medicine schools. So when I'm at Liverpool I'm collaborating with epidemiologists and public health people, whereas when I'm at Lancaster I'm training young statisticians.’

Promoting statistics as an integral part of science rather than only as a branch of mathematics is one of the key priorities that Peter has set out as part of his vision as president of the Royal Statistical Society - a role he will continue for the next two years having succeeded John Pullinger six months earlier than expected (John stepped down from the role after becoming the country’s National Statistician).

Peter joins a long line of distinguished statisticians who have served as RSS president over its 185 year history. ‘I can't think of one that I haven't admired,’ he says, when asked which president has inspired him the most. However, he picked out one of them - Maurice Bartlett - who was Peter’s supervisor in his first year as a research student. ‘I found him an extraordinarily kind person,’ he recalls. ‘He had a professional reputation for being rather remote and forbidding but my experience was that you couldn't ask him for the wrong thing. He was happy for me to get involved in what you might call applied, less mathematically heavy, things, and he encouraged me to go in that direction. I was only with him for a year but it was a formative experience.’

Peter was recommended to apply to study under Bartlett by Julian Besag, who had been a post-doc with Bartlett before becoming a lecturer in Liverpool where Peter was studying for his BSc. ‘Julian was the one who encouraged me to aim for an academic career rather than training to be an actuary - which may or may not have been just as interesting; I'll never know - and he suggested that I go and work with Maurice Bartlett.’

After gaining a lectureship at Newcastle upon Tyne and finishing his PhD, Peter was granted a sabbatical and went to work for six months in a rural forestry research station in central Sweden. ‘That was the first time I got exposed to real data,’ he remembers. ‘I really enjoyed it. Towards the end of my time in Sweden I did some field work. With two colleagues, I collected an original set of data (mapping the spatial pattern of heather) and realised that a lot of what I'd been doing earlier was sort of missing the point. In particular, I remember producing an analysis which was technically correct, but irrelevant to anything anyone at the forestry research station cared about.’

In 1978 Peter got a letter from CSIRO in Canberra, asking if he would like to join a programme of visiting lecturers there. He ended up working there for five years during the 1980s. ‘I learned so much at CSIRO, it made me really understand what data was about.’ he says. Back then, many of CSIRO’s statisticians maintained two offices: one in their statistical base, the other in the scientific division where they carried out their applied work. Peter remains convinced that this is the best way to foster interactions between theoretical and applied work. ‘You only achieve long term benefit if statisticians and scientists operate as wholehearted collaborators,’ he says. ‘In some medical schools, statisticians are still viewed as people whose role is to service the needs of the clinical researchers. Fortunately, this is less so now than it would have been a generation ago but it can still happen.’

Peter turned to medical and health research when he returned to England and in 1988 started at Lancaster University as professor of statistics where he was to go on to set up its Medical Statistics Unit in 1995. He also began collaborating with Scott Zeger, from the John Hopkins Bloomberg School of Public Health in the US. The partnership proved to be a fruitful one, and they worked together on a number of papers and a book, Analysis of Longitudinal Data, co-authored with Hopkins colleagues Kung-Yee Liang and, for its second edition, Patrick Heagerty

He admires the culture at John Hopkins, which he sees as being similar to CSIRO. ‘You've got world class substantive scientists and world class statisticians coexisting and collaborating with mutual respect,’ he says. ‘Statistics is not seen as a service role, it's an even-handed collaboration.’

The situation has improved in the UK, however. ‘Some of the most innovative statistics is now being done in departments which sit in medical schools or health faculties. It’s much better than it used to be. Social science now, equally, recognises that statistical thinking is integral to doing good social science. Both the Medical Research Council and the Economic and Social Research Council are very supportive of statistics.’

A different trend which has surged upwards in statistics during Peter’s career is the popularity of ‘Bayesian’ statistics. Does Peter consider himself a ‘Bayesian’? ‘Well,’ he replies, ‘you can't not believe in Bayes’ theorem because it's true. But that doesn’t make you a Bayesian in the philosophical sense. When people are making personal decisions - even if they don't formally process Bayes’ theorem in their mind – they are adapting what they think they should believe in response to new evidence as it comes in. Bayes’ theorem is just the formal mathematical machinery for doing that.’

He is, however, less convinced that this approach works best in science. ‘My fundamental concern is that if you really follow the Bayesian paradigm philosophically, you admit that different scientists will interpret the same evidence differently. For making personal decisions, that's entirely legitimate. But in terms of trying to find out what nature is up to, it’s as (US physicist) Richard Feynman famously said: “Nature cannot be fooled”. I don't fundamentally buy the Bayesian philosophy as a basis for establishing scientific truth in the way that I do for intelligent personal decision-making.’

‘I think a lot of people have become Bayesians simply because they can be, using modern computational methods.’ Peter continues. ‘And that's fine. But it does mean that a lot of what's published, I think, has within it wrinkles that are hidden by the elegance and the simplicity of the Bayesian formalism. So while people can easily check that their main conclusions are not heavily influenced by pretending to change their prior beliefs, there are subtle aspects that they can't check. I think it’s too glib to say that because Bayesian methods are elegant and beautiful they’re necessarily the right tools to use in all circumstances.’

It is a debate that is likely to continue both within and outside of the Royal Statistical Society’s academic journals and ordinary meetings. Peter has had a longstanding relationship with the Society and has held many roles within it since becoming a member early on in his career. ‘Back then, I didn't know any research active academic that wasn't a member of the RSS,’ he says. ‘You didn't even think about it, you just did it. That was how you kept in touch with what was going on - by reading the journals and going to the meetings. There was no other way - no internet, no websites.’

When Peter became honorary secretary some years later (1990-96) he oversaw the Society merge with the Institute of Statisticians. ‘Before the merger it was a scholarly society, simply run,’ he recalls. ‘It was clear that in merging, we were going to become a society with a dual role - a scholarly society and a professional body, which has a role in influencing public policy and making sure that the value of statistics is fully recognised and that statisticians have proper professional support.’

Another way of promoting statistical awareness in society is to promote it within secondary schools, which is the second of Peter’s key priorities as president. The new post-16 core maths qualification, into which the RSS has had significant input, is an exciting development in this area. ‘Core maths means that it's not just a specialist thing,’ he says. ‘It's like core English - everybody needs to be literate and numerate. That's the message.’

He would also like to see more informal activities promoting statistics in secondary schools, such as the Guy Lecturer. ‘I would like to see that become a Guy team - a dozen people going into schools,’ he adds.

Peter’s third priority as president is doing more for and with developing countries, having worked on research programmes in northern Brazil and various countries in Africa. He recalls being ‘overwhelmed’ by the social problems in those countries, albeit coupled with an ‘enormous warmth of welcome’ that he has received there. ‘It just personally hit home that maybe I could in a small way contribute to something that dwarfs everything else in importance,’ he says. ‘The world is not in a good state and in a small way we should do more for, and with, the poorest countries in the world. The Royal Statistical Society has a couple of vehicles which we can do this with: one is the pro bono work (the Statisticians for Society initiative), the other the International Development Working Group.’

Peter is also very supportive of the Society’s Young Statisticians section. ‘Young statisticians have always been part of the Society, but it seems to me that these younger people have a huge amount to contribute, not just as members of the Society but as decision makers within policy setting, honorary officers - the works,’ he says. ‘The youngsters make me really optimistic about the future of the Society.’

Health & Medicine Peter Diggle Young Statisticians Section (YSS) Bayesian statistics