Please note that due to her absence from the UK from September 23rd, the Regional Treasurer, Gillian Arnold, has stood down a few weeks in advance of the AGM in November. In the period prior to the election of a new Regional Treasurer at the AGM, any financial matters should be referred to the Regional Secretary, Andrew Mead (contact details below).
¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤
Following the successful Regional Conference (see later), this year the Autumn Meeting of the British Region of the International Biometric Society will just be a half-day meeting, incorporating the Annual General Meeting. This meeting will be held at 2pm on Thursday 13th November 2003 in the Lecture Hall at the Royal Statistical Society, 12 Errol Street, London, EC1Y 8LX .
The programme for the meeting is given below, together with short abstracts for the papers to be given during the meeting. The Agenda for the Annual General Meeting follows at the end of the programme. There is no charge for this meeting and no need to register in advance.
Information on how to get to the Royal Statistical Society can be obtained from their web-site http://www.rss.org.uk/about/direction.html.
14:00 – 14:30 Annual General Meeting (for IBS British Region Members only)
14:30 – 15:30 Biometrical issues in the Farm Scale Evaluations of Genetically Modified Herbicide-Tolerant Crops
| 1Joe N. Perry, 2Peter Rothery & 1Suzanne Clark |
| (1Rothamsted Research & 2Centre for Ecology & Hydrology) |
A review will be given of the challenges in design and analysis of the Farm Scale Evaluations of Genetically Modified Herbicide-Tolerant (GMHT) crops. This experiment has been conducted by a team of academic ecologists from the Centre for Ecology & Hydrology, Rothamsted Research & the Scottish Crop Research Institute. It is funded by Government to study the effect on farmland wildlife of two contrasting herbicide management regimes: one from conventional agriculture and one arising from the use of GMHT varieties. If the results from the spring crops are already in the public domain, the review will be followed by a summary of the results, and a discussion of the opportunities for further work.
Two methods papers were published in February 2003, in the Journal of Applied Ecology, one biological, and one largely concerned with biometry. In addition, an invited paper was given at the 2002 Freiburg IBC, and a written version of this is in press for Environmetrics. At the time of writing this abstract no results have yet been published, but there have been some recent criticisms of our methods made in the media and elsewhere.
These centre on:
This is an experiment of national importance, that raises in addition other more technical statistical questions concerning the discrete nature of the count data collected, which often comprise a large proportion of zeroes.
15:30 – 15:50 Tea
15:50 – 16:35 Bayesian estimation for percolation models of disease spread in plant communities
| Gavin Gibson |
| (School of Mathematical and Computer Sciences, Heriot-Watt University) |
There is a growing body of research on fitting spatially explicit models to data on the spread of plant disease. Some of this work has focussed on developments of the classical epidemic models such as the SI or SIR formulations. In this talk we describe the development of percolation-based models of epidemic spread. Specifically, we consider a disease which spreads through a host population whose members are located at the vertices of a lattice, via a combination of transmission mechanisms. These include infection from primary inocculum, secondary plant-to-plant transmission and the potential change in the host susceptibility as the epidemic proceeds.
The main part of the talk describes how such models can be fitted in a Bayesian framework to data describing disease status at (possibly sparse) sample times. This is done by formulating MCMC algorithms to explore the joint posterior distribution of model parameters and hidden aspects of the process. In particular, we show how inferences can be made on the hidden chains of infection that have occurred in the population. The approach is applied to the study of fungal diseases in radish and the balance between primary and secondary sources of infection is explored for this system.
16:35 – 17:20 Spatially varying coefficient models in ecological studies of environment and health
| Nicky Best, Sara Morris, Sylvia Richardson |
| (Department of Epidemiology & Public Health, Imperial College School of Medicine, UK) |
Effect modification is recognised as an important phenomenon in epidemiological studies and occurs when the exposure-response relationship depends on the level of a third variable (the effect modifier). Effect modification is common at the individual level - for example, age or genetic factors are known to increase individual susceptibility to disease due to certain risk factors. However, effect modification may also occur at a group (area) level due to e.g. contextual effects (say, living in a poor area) or the effectiveness of the health care system. In this talk, we discuss the use of varying coefficient regression models to capture geographical heterogeneity in the effects of risk factors on disease outcomes. We consider small-area ecological regression models, which involve fitting a Poisson model to counts of disease cases in a set of contiguous geographical areas and modelling these as a log-linear function of mean exposure in each area. Spatially structured random effects are generally included to account for the effects of unmeasured risk factors that may vary smoothly across space.
Varying coefficient models replace the constant regression coefficient by a separate coefficient for each area, the idea being to capture the interaction between the effect of exposure and location (where the latter represents a proxy for the effect modifier). One difficulty with such models is that the varying coefficients are not identifiable unless repeated measurements are available for each area. This is rarely the case, and so we instead make use of our prior beliefs that the effect modifier is likely to vary quite smoothly across areas and so model the regression coefficients as random effects with an appropriate spatially-structured prior distribution.
This talk will discuss various Bayesian models for spatially varying ecological regression coefficients and explore their sensitivity for detecting different patterns of ecological association. The approach will be used to study the association between exposure to environmental benzene emissions and risk of childhood leukaemia in London.
17:20 Close
¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤
The Annual General Meeting of the British Region of the International Biometric Society will be held at 14:00 on Thursday 13th November 2003 in the Lecture Hall at the Royal Statistical Society, 12 Errol Street, London, EC1Y 8LX .
The following have been proposed by the Nominating Committee:
| Dr. Steve Brooks (University of Cambridge) |
| Mr. Peter Colman (Pfizer Global R & D, Sandwich) |
Continuing in office:
| Dr. Kate Tilling (University of Bristol) |
| Mr. Dave Waddington (Roslyn Institute) |
| Prof. Sarah Darby (Cancer Research UK, CTSU, Oxford) |
| Dr. John Whittaker (Imperial College) |
Any two members of the region may make additional nominations. To be valid, such nominations shall: a) be in writing; b) reach the Regional Secretary by October 16th (four weeks before the Annual General Meeting), and c) have the consent of the person(s) nominated.
Copies of the minutes of the last AGM and of the reports to be presented at this meeting will be posted on the region's web site (address below) approximately four weeks before the meeting.
http://www.maths.qmw.ac.uk/~rab/biometrics/british.html
¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤
¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤
¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤
¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤
Andrew Mead