Biomathematics and Statistics Scotland
JCMB, The King's Buildings,
Peter Guthrie Tait Road,
EDINBURGH, EH9 3FD, Scotland, UK.
Tel: +44 (0)131 651 7032
Email: Chris Theobald
» DesignSequence Page
I had a joint appointment in BioSS and the School of Mathematics, University of Edinburgh from 2001 to 2011. I now contribute to external projects on potato trials, insect detection in salads and optimum nitrogen fertilization.
My main interest is in the application of Bayesian methods in agriculture. I argue that Bayesian
procedures should replace conventional estimative approaches in order to make systematic use of expert knowledge.
These are some applications of my ideas.
- Crop varieties for local environments With Mike Talbot of BioSS I have investigated the
application of Bayesian decision theory to the selection of crop varieties best suited to local environments.
Funding from the Home-Grown Cereals Authority (HGCA) has supported the development of computer-based
tools to help farmers and advisors identify varieties best adapted to local circumstances: these are available
- Fertilizer optimization We have also used decision theory to find the optimum choice of variety and
fertilizer level for cereals, taking account of crop yield and fertilizer cost. We have since extended our method to allow
the value of the crop to depend on a quality measurement, such as grain nitrogen, and
to include measurements of soil nitrogen.
- Seed rate optimization With Adrian Roberts of BioSS and funding from HGCA, I have
investigated optimum seed rates for winter wheat, and how they should vary with sowing date and latitude over the UK.
- Predicting fungal contamination in wheat Contamination of wheat grain by bunt (Tilletia caries) is usually measured by microscopic counting of spores washed from a seed sample, but polymerase chain reaction (PCR) technology offers a quicker and more convenient alternative. A Bayesian calibration model developed with Adrian Roberts accounts for sampling variation in both methods, and bias, censoring and run-to-run variation in the PCR method.
I am also interested in the design and analysis of visual perception experiments. With Aletta Nonyane and
the late Rob Kempton, I have investigated the construction and properties of designs for visual perception experiments
which are balanced for carry-over effects. These designs are now being used in neuroimaging. See the DesignSequence Page for more information. We have also developed and applied a Bayesian predictive method for
correcting bias in sequences of quantitative visual responses.
With Ayona Chatterjee and Graham Horgan of BioSS, I have considered Bayesian models for dietary data recorded on successive days, particularly components of diet which have very skewed distributions and large proportions of zeros, such as retinol, alcohol and pesticide residues.
We have developed a multivariate latent-Gaussian model for the consumption of Iprodione, a widely used pesticide, combining information over several food products on which it is used.
Another interest is in multivariate analysis, particularly discriminant analysis and pattern recognition.
- Automated recognition of chromosomes and cereal seeds With Andrew Carothers of the
MRC Human Genetics Unit and Simon Kirby I developed several discrimination methods for improving the
speed and accuracy with which chromosomes could be recognised. I have since supervised research on the automated
recognition of cereal seeds, in cooperation with the Scottish Agricultural Statistics Service and
the Scottish Agricultural Science Agency.
- Automated segmentation of CT X-ray images With Chris Glasbey of BioSS and Caroline Robinson I
worked on methods for automated segmentation of computed-tomography X-ray images of sheep.
Considering how to exploit the approximate bilateral symmetry of the X-ray images led us to investigate
principal component analysis of landmarks from what we call reversible two-dimensional images.
Journal Publications from 2002
(available as pdf)
- Estimation of economically optimum seed rates for winter wheat from series of trials. J H Spink, M Talbot, C M Theobald and A M I Roberts (Journal of Agricultural Science 144 (2006), 303-316)
- Bayesian solutions to some decision problems in crop management and variety choice. C M Theobald (Plant Breeding Lecture Series, Iowa State University (2006))
- Assessment of expert opinion: seasonal sheep preference and plant response to grazing. M L Pollock, J P Holland, C M Theobald and C J Legg (Rangeland Ecology and Management, 60 (2007), 125-135)
- Calibration of quantitative PCR assays. A M I Roberts, C M Theobald, and M McNeil (Journal of Agricultural, Biological, and Environmental Statistics 12 (2007), 364-378)
- Design sequences for sensory studies: achieving balance for carry-over and position effects.
B A S Nonyane and C M Theobald (British Journal of Mathematical and Statistical Psychology, 60 (2007), 339-349)
- Multiplicative models for combining information from several sensory experiments: a Bayesian analysis. B A S Nonyane and
C M Theobald (Food Quality and Preference, 19 (2008), 260-266)
- Exposure assessment for pesticide intake from multiple food products: a Bayesian latent-variable approach. A Chatterjee, G W Horgan and
C M Theobald (Risk Analysis 28(2008), 1727-1736)
- A hierarchical Bayesian mixture model for repeated dietary records.
C M Theobald, A Chatterjee and G W Horgan (Food and Chemical Toxicology 50 (2012) 320-327)
- Group testing, the pooled hypergeometric distribution, and estimating the number of defectives in small populations. C M Theobald and A M Davie (Communications in Statistics Theory and Methods 43 (2014) 3019-3026)