Over 100 biometricians (including 27 past Reading students) came to Reading for the British Region conference, 16-18 September 2003. Invited and contributed talks and posters covered the whole range of biometry today. It was simulating to see the common statistical threads which unite these apparently diverse research areas and how insights gained in one area can be used in another.
Scott Zeger (Johns Hopkins University) opened the conference with an accessible presentation "Does the air we breath kill? Evidence from spatial-time-series analyses of the National Morbidity and Mortality Air Pollution Study (NMMAPS)". After a review of the evidence on the health effects of air pollution, he discussed the differences between cohort and time-series methodologies for estimating these effects. He described the analysis of NMAPS (a time-series study), and then went on to describe how such studies could be connected with cohort studies.
Unfortunately there is not space to give details of all the talks. The programme consisted of five invited sessions, ten contributed sessions (36 papers) and 13 posters. However, the programme can be viewed on the regional web-site. Two of the highlights were sessions organised in honour of the 70th birthday of Professor Robert Curnow and the 65th birthday of Professor Roger Mead. Each session began with a presentation by a former PhD student.
Richard Mott (Wellcome Trust Centre for Human Genetics, Oxford) opened the session in Robert's honour with a presentation on finding genes affecting quantitative traits in mice. After outlining why the mouse is an ideal model for genetic dissection of quantitative trait loci (QTL), he outlined the statistical and bioinformatics issues in fine mapping QTLs. For one QTL under consideration over 1000 polymorphisms that could give rise to the trait have been identified. Richard described statistical measures that could be used in such situations to eliminate the majority of candidate polymorphisms from consideration.
Shelia Bird (MRC Biostatistics Unit, Cambridge) continued the session with her talk "Prions and Projections: from vCJD to Hepatitis C". After paying tribute to the "Curnow Trio" (3 papers in Applied Statistics on the BSE maternal cohort study looking at direct maternal transmission versus inherited genetic susceptibility for increased BSE risk to calves), Sheila discussed the issues in projecting the number of future cases of BSE and vCJD. Beginning with a review of recent back-calculations of the UK BSE epidemic, which suggested previous estimates underestimated the cases fourfold, she went on to discuss recent elegant work which successfully explains differing incidence of vCJD by birth cohort and sex to using information on different dietary exposure patterns and age-susceptibility. Turning to Hepatitis C, she described how pieces of the projection jig-saw relating serious liver sequelae to injector-related Hepatitis C transmission dynamics, mortality from other causes, treatment uptake and other covariates are being fitted together. Sheila concluded by calling for a new Curnow trio, in the form of working parties on drug-related deaths, hepatitis C projections and criminal justice interventions for drug-dependent offenders.
Steve Gilmour (Queen Mary College, London) opened the session in honour of Roger Mead by speaking about open design problems in complex experiments, and exploring how recent trends towards increasingly complex treatments and responses, together with sophisticated statistical modelling, fitted in with randomisation and design. Arguing that there were many new and challenging design problems to be tackled, he highlighted the importance of considering experimental and sampling design, together with any mechanistic information, to ensure experimental data contain the required information.
John Eccleston (University of Queensland, Australia) continued the session with a talk on multi-environment and two-phase experiments. The field trials he described consisted of hundreds of treatments with many different sites and spatial dependence within sites. He described two optimality criteria (A-optimality and N-optimality) which aim to allocate treatments to plots to get the most efficient estimates, and illustrated how N-optimality, which is several orders of magnitude faster computationally, provides good approximations to A-optimality in practice. He then went on to describe how wheat from such trials went into a second phase of analysis, where its milling, dough making and other qualities were examined. However, optimal design for the first phase does not ensure optimal design for the second. Further, the second phase of the experiments currently involve little design and modelling. John explored some initial ideas for these interesting and challenging issues.
The excellent conference atmosphere was epitomised by the conference dinner, held in the panelled dining hall in Wantage Hall. After dinner, Rosemary Bailey (regional vice president) presented the Cedric Smith Prize (sponsored by Wiley) to Suzanne Clark (Rothamsted Research) and colleagues for their poster "Modelling spatial trend in aphid suction trap data at the European scale". After a nasty moment when Professor Sir David Cox threatened post-prandial proceedings with the details of second order ancillarity, he instead treated us to entertaining reminiscences of some of those involved in the early days of the International Biometric Society, both in the UK and USA.
The final session was a whistle-stop tour of recent developments with randomisation and other computer-intensive methods in Biology from Bryan Manly (Western Eco Systems Technology, Wyoming, USA). Bryan delineated the differences between randomisation, bootstrap and MCMC methods and gave examples of each in action. The ability of randomisation tests to provide tailor-made solutions was illustrated by an "empty top-left corner" test for the association between rainfall and year in a scatter-plot. Bootstrap methods were described for estimating the uncertainty of parameter estimates in two-phase adaptive stratified sampling and a Monte-Carlo method for density estimation in line trapping studies was outlined.
The conference concluded by expressing the British Region's gratitude to all those who contributed to the meeting's success, particularly the local organising committee (Chair Derek Pike) and the programme committee (Chair Andrew Mead). The next regional conference is already being planned for April 5-7 2005 at the University of Leicester. Book the date!
Report by James Carpenter, British Regional corespondent, September 2003.