In celebration of the International Year of Statistics, we reached out to a few British statisticians to ask them to share a few thoughts on statistics as a discipline, statisticians and applied statistics. The United Kingdom has a rich statistical history, and we are pleased to share some of these most interesting blog interviews.
Our first interview is with Dr. Shirley Coleman, Technical Director and Principal Research Associate at Newcastle University (NU). She is an applied statistician with wide-ranging experience in manufacturing, sales, finance, planning and healthcare, drawing on a variety of techniques to deliver effective solutions. She is widely published and very active in the greater statistical community over many years — Director of Industrial Statistics Research Unit at Newcastle University, President of the European Network of Business and Industrial Statistics, Chair of Quality Improvement Section for the Royal Statistical Society (RSS) — to highlight a few of her many undertakings to promote the discipline of statistics.
1. What was your reaction when you first heard about celebrating the first-ever International Year of Statistics in 2013?
I don’t like the concept of “year of,” so I wasn’t particularly impressed by the idea of the “International Year of Statistics.”
2. Though Statistics2013 is a yearlong celebration, the GetStats campaign in the UK is a 10-year effort (we think that's awesome!). Two years into it, how do you think it's going?
GetStats is a great campaign. It has already has had some brilliant impacts. In particular, it has enabled statisticians to reach Members of Parliament and build on the RSS work with journalists. It is well-established in place at the centre of a much wider trend towards raising the profile of numeracy. It works as a catalyst for networking, for example ongoing discussions with the Open Data Institute regarding our mutual aims.
3. Who are your favourite statisticians and why?
Amongst my favourite statisticians are three of the past professors of statistics with whom I worked at Newcastle University, now sadly passed away. They are all very different people. Professor Robin Plackett was fascinating; I once interviewed him about his time in an elite wartime group of fresh Cambridge graduates who were tasked with tackling operations research-type questions for the war effort, looking at such things as the number of vests to send out to troops to ensure an adequate supply. Robin was a stately, modest person who was revered by his many PhD students and well-respected by his colleagues. He produced his paper on the Plackett Burman experimental design in 1946 when a young lecturer. He announced his intention to retire when he reached 65 at the end of the academic year, just before the frightful news that Professor John Anderson had collapsed and died in the sports hall at the age of 43. Robin was asked to stay on for another year but declined, saying he was not irreplaceable. John was an expert in multivariate analysis; he told us that as a youngster he had been ill and confined to bed for a few weeks, and to pass the time, he read a book on linear algebra and acquired the expertise from which sprung his subsequent brilliant work. Robin Plackett read John’s paper at the RSS word for word from John’s manuscript as he considered it not his place to embellish or change it. John was a shining light and intensely lively and popular. I was lecturing at NU in 1983 during the year that he died. We had a fire alarm in my first lecture, and John sent me a warm note hoping that it hadn’t put me off. When I told him about a job I wanted but was unsure if I was good enough for it, he told me that as a principle one should always apply for any job one fancied regardless, and that is extremely good advice. Professor Julian Besag was a wild character, and he was brutally intelligent and direct, with a loyal coterie of admirers. He was a glamorous figure and did great work in Bayesian analysis and spatial statistics.
4. What are some of the things you most appreciate about statistics in everyday life?
For me, statistics crystallises clarity from chaos. Hearsay and subjectivity melt under statistical scrutiny, and that is extremely satisfying. For example, a friend of mine has a business and a penchant for figures; she gave me her daily retail data for the past year: numbers of shoppers entering her shop, the daily takings, the weather and the main salesperson. From this data, statistical analysis reveals that the proprietors extract more sales per shopper than other employees; the income per hour during the week is fairly constant, so the opening hours are well chosen; income per shopper is higher on poor weather days, etc. On the other hand, I am wary of how statistics can oversimplify things and strip out the wealth of context, for example, questionnaires and decision matrices in which correlations are guessed as strong, medium or weak and importance scores are weighted seemingly at random.
5. What areas of statistics do you think have much to offer yet are underutilized?
Statistical process control (SPC) is reasonably well used but still has much to offer in new contexts. Design of experiments (DOE) could be better used, although applying factors can lead to the trap of oversimplification. It would be good to see DOE combined with SPC more. For example, an SPC chart shows a decline in service or performance; a team brainstorms an action plan and adopts it, and then marvels at the improvement as evidenced by the SPC chart. However, this is an all or nothing one-factor experiment, and it would surely be beneficial to identify the factors in the action plan and try a designed experiment with a few trials that could demonstrate which of the factors or their interactions were responsible for the improvement.
6. What other thoughts about statistics would you like to share?
Statistics as a profession has an interrelated upside and downside. As a subject that can be applied widely, it enables the statistician to be involved in all sorts of interesting research and analyses; the corresponding downside is that the statistician is not core to the research. This was illustrated to me when the different groups at NU were invited to discuss their consultancy income. The agricultural group who presented first had income in £millions. Unfortunately, I had to follow that presentation, and really statistics income is so much lower. This is partly because we do not have equipment or facilities to share, but it is also because we do not have a core expertise that is unarguably needed like expertise in deep-sea drilling or organic farming. Our expertise is fundamental but is usually sold to other researchers and forms an inextricable part of their product. Trying to sell statistics consultancy is much harder than selling materials or machinery expertise. A geology student warned me of this when I was a student; he pointed out that whereas his specialty was always recognisable as a key subject in itself, statistics was always going to be a contributor. Nevertheless, statistics is interesting, and it appeals to me.
Thanks to Shirley for taking the time to share her interesting perspective and help celebrate Statistics2013. (We hope to extend the momentum beyond the current year as we agree with her “year of” doesn’t take the long-term view. We should continually promote the discipline of statistics, statisticians and the effective application of statistics!) Look for chats with UK statisticians coming in the JMP blog.
 Diggle, Peter and Henderson, R. and Matthews, J (2010) Obituary : Robert Lewis Plackett, 1920-2009. Journal of the Royal Statistical Society, Series A, 173 (1). pp. 259-267.
 Plackett, RL, and Burman, JP (1946), “The Design of Optimum Multifactorial Experiments,” Biometrika, 33, 305-325 and 328-332.
 Obituary (1984) in JRSS, 147(1), 120.
 Anderson, JA, (1984) Regression and ordered categorical variables, JRSS(B), 46(1), 1-30.
 Diggle, Peter and Green, P J (2011) Obituary : Julian Besag, FRS, 1945-2010. Journal of the Royal Statistical Society, Series A, 174 (2). pp. 502-504.
 Besag, J and Green, PJ (1993) Spatial statistics and Bayesian computation, JRSS(B), 55(1), 25-37.