International Biometric Society
British Region

205^th Ordinary Meeting
In memory of Rob Kempton
Friday 16^th April, 2004, 9.30 - 17.30
Royal Society of Edinburgh

The 205^th Ordinary Meeting of the British Region will be a special meeting held in memory of Rob Kempton, past President and Secretary of the Region, who died in May 2003, shortly after election as President-Elect of the International Biometric Society. This one-day meeting will be held in the Wolfson Lecture Theatre of the Royal Society of Edinburgh, of which Rob was a Fellow, starting at 9.30 (registration from 9.00) and finishing at about 17.30.

We hope that many members will want to, and be able to, attend this special meeting.

It will be necessary to register in advance for this meeting using this registration form, at a cost of £35 for IBS members and £45 for non-members.

We have also reserved some accommodation for the Thursday (and Friday) night in the Kenneth Mackenzie Hospitality Centre (see below) and arranged a meal for those present in Edinburgh on the Thursday evening. A list of B&Bs and Hotels is also available here.

The full programme for the meeting follows.

Rob Kempton (1946 – 2003)

After attending Chislehurst and Sidcup Grammar School, Rob read mathematics at Oxford University, graduating from Wadham College in 1968. He followed this by a B. Phil. in Applied Statistics, an almost unique qualification as the course was discontinued after just one year.

Rob's first job, between 1970 and 1976, was as a statistician at Rothamsted Experimental Station. He made major contributions from the start, including work on the Maximum Likelihood Program (MLP) with Gavin Ross. He set the pattern for a successful liaison statistician scheme by which a statistician is attached to a specific department and he co-authored diverse papers for leading ecological and statistical journals, as well as the book with PGN Digby on "Multivariate analysis of ecological communities". His work was well ahead of its time and Rob's writing demonstrated an incisive style, both sound mathematically and with understandable messages for the scientists

In 1976 Rob was appointed Head of Statistics at the Plant Breeding Institute in Cambridge, where he worked for the next ten years. His subsequent contributions to research in the design and analysis of experiments with carryover of treatment effects between experimental units was stimulated by his observations of competition between adjacent varieties in plant breeding trials at Cambridge. The book that he edited with PN Fox, "Statistical methods for plant variety evaluation", encapsulated, in Rob's typically clear and concise style, many of the statistical good-practices in plant breeding.

Rob moved to Edinburgh in 1986 to take over as director of the newly-created statistical consultancy and research organisation that came to be known as Biomathematics and Statistics Scotland (BioSS). When appointed Director of SASS, the predecessor of BioSS, he had the difficult tasks of forging relationships between his new organisation and the various other Scottish agricultural research institutes, and of inspiring a team of statisticians whose future had seemed uncertain. He did all of these things brilliantly, as the recent record of BioSS shows, and his management of BioSS has shown how a statistical research and consultancy service can flourish.

Rob was an enthusiastic supporter of the International Biometric Society (IBS) for many years, serving in many roles. He was proud to be President-elect of IBS and, at the time of his death, was already making plans for his presidential address, as well as developing many ideas for the future development of the IBS both internationally and within the British Region. He gave many years service to the work of the IBS, first serving on the British Region committee for three years from autumn 1980. He was elected Regional Secretary in autumn 1983, serving in this office for four years, and served as Regional President between autumn 1994 and autumn 1996, continuing to serve on the regional committee for the following two years as past-President. He was an active President of the British Region, finishing with a memorable presidential address on `Life in a competitive environment' in October 1996, and then leading the organisation of the successful conference to mark the 50^th Anniversary of the British Region in Edinburgh in 1998. He also served two four-year terms as an IBS Council member (1984 – 1987 and 1990 – 1993). Perhaps his greatest personal contribution was in helping develop the Sub-Saharan African Network (SUSAN), an activity in which he successfully combined his professional talents with his ideals of social justice.

In that spirit Rob played a crucial role in having the IBS meet in Africa for the first time in 1998. The sub-Saharan network meetings arose out of his enthusiasm and the financial support he generated. Rob led by a quiet example of rigorous professional excellence, and a generous and open spirit. He combined a ready sense of humour and a diplomatic style that built bridges where there had been none. He reached out from his first activities in central Africa, to north and south, east and west and connected a growing group to the IBS, its global community, and resources. He clearly communicated that biostatistics has much to contribute to the definition of food, agriculture and health imperatives, and has the potential to transform the lives of the poor and forgotten everywhere. Somehow he undertook these tasks with a quality that might well be described as joyous. Later, he helped in the formation of a similar network in the Caribbean. He had been elected as International President-Elect at the start of 2003, and had already started helping to shape the future of the society before his untimely death.

In the last few years he represented the British Region of the International Biometric Society on the Fisher Memorial Trust, organising the 2002 Fisher Memorial Lecture given by Oliver Mayo at the Royal Society of Edinburgh. He also served on numerous committees and review groups, most recently for the Food Standards Agency, and was establishing himself as an expert on the statistical analysis of risk.

This meeting provides an opportunity to reflect on developments in the many and varied areas in which Rob worked, and on Rob's contributions in these areas. We also pay tribute to his contributions to biometry in the UK, in particular his founding and directing of BioSS, and to his enthusiastic support of the International Biometric Society.

Programme

09.00 Registration

09.30 Tribute: Rob and Biomathematics and Statistics Scotland

10.00 Tribute: Rob and the International Biometric Society

10.20 The statistically diverse environment

Within environmental and ecological research diversity, multivariate responses and lack of replication are constant reminders that the natural environment is rather variable and is often resistant to the imposition of simple concepts from statistical research. Rob Kempton was active in developing diversity statistics 25 years ago and subsequently published a book on multivariate analysis. Both of these activities, along with his work on local spatial effects, have links into recent advances in understanding of ecological and environmental issues.

11.05 Coffee

11.25 Design of field experiments in the presence of interference between treatments

In many types of field experiment it is likely that the treatment applied to one plot will affect neighbouring plots as well as the plot to which it is applied. Such an effect is variously called a neighbour effect, or interference between treatments, or competition between treatments. Rob Kempton was among those who recognised such effects and the need not only to model them in the analysis but also to design the experiment with the neighbour effects in mind. This talk will give recommendations for design, some derived theoretically, some more ad hoc, together with examples from various different types of treatment.

12.10 Crop cultivar breeding and evaluation: a review of current statistical methods

In order to meet the challenge of feeding a rapidly increasing world population it is crucial that food production be increased. An important means by which this an be achieved is the breeding and subsequent adoption of new crop cultivars that are higher yielding, resistant to a range of diseases and have superior quality in terms of end-use product manufacture (e.g. bread making). For this reason many countries have a substantial commitment to plant breeding and cultivar testing programmes. The key goal of such programmes is the selection of superior cultivars, either for continued testing (in early stages of breeding programmes) or for recommendation for commercial use. Biometricians have a key role in this process, namely to ensure that selection errors are minimised. Although this goal is universal, the statistical methodology used may vary. In this paper we review current methodology. We focus on our Australian experience and the UK system into which Rob Kempton had significant input.

12.55 Lunch

13.55 Applications of Markov random fields in biometry

Despite their flaws, Markov random fields have been used with some success in accounting for spatial variation in several different areas of biometry. This talk will briefly describe the underlying ideas and provide some examples, drawn from agricultural field experiments, where my first paper was joint with Rob Kempton, from spatial epidemiology and from image analysis of gene expression data.

14.40 Modelling neighbour effects with plant height in variety trials

Neighbour effects between plots may bias variety estimates in variety trials. Bias due to neighbour effects can effectively be reduced by using plots with unharvested border rows. When this solution is not practicable, bias may be reduced by design and statistical analysis. This talk will deal with statistical methods which are appropriate when neighbour effects are related to plant height. It will consider Rob Kempton's contribution to the area, nonparametric tests of hypotheses on neighbour effects, optimal designs for covariance analysis, and a bivariate modelling of yield and height.

15.10 Analysis of sickness absence data

Many organisations record sickness absence to ensure that employees who are absent from work due to ill-health are treated fairly and are provided with support towards full recovery. Monitoring of sickness absence enables management to identify major factors associated with sickness absence and provides necessary information for human resource management. Employees with unsatisfactory sickness absence record can be identified and helped to improve their attendance record.

One of Rob Kempton's research interest was to provide sound methodological framework for describing and making statistical inference on the sickness absence phenomenon. Rob Kempton's last major statistical undertaking was a project to analyse sickness absence data for the Scottish Executive compiled by BioSS for the period 1993-2000. In this paper, I present some of the work done in collaboration with Rob Kempton in that project. I describe statistical models that account for the binary nature of the data to model variation in absence levels over time.

I will also highlight Rob Kempton's role in the development of Biometry in Africa.

15.40 Tea

16.00 Multivariate analysis of ecological communities

In this talk I review some of the developments since Rob Kempton's book "Multivariate Analysis of Ecological Communities" (Digby & Kempton, 1987). In my view niches and unimodal models are the key for understanding ecological data. This point of view led to a mathematical underpinning for the use of correspondence analysis for ecological data and to the development of constrained correspondence analysis methods

1. to relate species data to environment data - (partial) canonical correspondence analysis, which integrates multiple regression and correspondence analysis in the same way as least-squares reduced rank regression (alias redundancy analysis) integrates multiple regression and principal component analysis
2. to infer environment data from species data - WA-PLS - and, recently,
3. to relate two species data sets - co-correspondence analysis, which integrates CA and multivariate PLS.

Statistical rigour of these crude methods is obtained by cross-validation and (approximate) permutation, also for situations where there is no strict exchangeability of the statistical units. These permutation tests are a prominent feature of Canoco for Windows (www.canoco.com).

The trend is towards explicit models, such as Yee and Hastie's (2003) Reduced Rank Vector Generalized Linear Models, and Bayesian analysis, such as in modelling metapopulation dynamics to estimate a species' extinction probability and in latent class approaches to cluster analysis to improve upon the good-old TWINSPAN method.

16.45 Estimation and Presentation of Risk

The estimation and presentation of risk has been an important but often contentious area of research and debate over recent years. Rob Kempton's recent work has been on some of the most important and difficult areas of risk assessment. In work for the Food Standards Agency on the risk assessment of mixtures of pesticides, Rob has stressed the importance of probabilistic modelling incorporating the consequences of stochastic variation in individual consumption and of uncertainties in the estimation of parameters. Rob has argued for a move away from testing hypotheses of no risk to testing whether the risk is below some acceptable level and for the importance of attaching measures of uncertainty to all quoted estimates of risk. His work and related work on the measurement and the modelling of the consequences of aggregate and cumulative consumption of pesticide residues in food and his contributions to the debates about BSE, GM crops and GM foods will be described.

17.30 Close

¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤

Accommodation

(1) Provisional reservations for April 15^th and 16^th have been made at

This is a central location, just east of Nicholson Street, only a short walk north to the Royal Mile, and about 1 mile from both the Royal Society of Edinburgh building (22-26 George Street) and la Bonne Vie restaurant (49 Causewayside) to the south.

There are: 18 single rooms (£27.50, £29.50 or £31.50 per night)

(2) For those wishing to try elsewhere, Edinburgh Tourist Board has an excellent web-site with many accommodation options and accompanying maps.

A selection of options and links to maps from the above site can be found here. They are chosen to be within about a mile of the restaurant, and on or near major bus routes into the east end of the city centre.

Evening meal – Thursday April 15^th, 8 p.m.

A party of up to 30 people can be accommodated at la Bonne Vie Restaurant, 49 Causewayside, which does reasonable French style cooking. A 2-course meal will cost about £16, or slightly more. There is no alcohol licence, but diners are welcome to bring their own bottles.

Maps showing the location of the Royal Society of Edinburgh, The Kenneth Mackenzie Hospitality Suite and la Bonne Vie Restaurant are included below.