Peter Craig

	Position:	Senior Lecturer in Statistics
	Address:	Dept. of Mathematical Sciences, University of Durham, South Road, Durham DH1 3LE, England
	Phone:	(0191 or +44 191) 3343076
	E-mail:	p.s.craig@durham.ac.uk
	FAX:	(0191 or +44 191) 334 3051

Contents

• Research Interests
• Recent Publications
• Possible Topics for M.Sc. and Ph.D. Research
• Current Research Students

Research Interests

Since November 2004, I have been providing statistical advice to the European Food Safety Authority's Panel on "Plant Health, Plant Protection Products and Their Residues". This is fundamentally an inter-disciplinary activity and much of the interest lies in finding pragmatic solutions to difficult problems. I have worked on two scientific opinions on: (i) variability factors for use in assessment of risk to human beings from pesticide residues in fruit; and (ii) uncertainty factors for inter-species variability in sensitivity to pesticides in relation to risk assessment for new compounds. This work has involved collaboration with the members of the EFSA panel, in particular Andy Hart who leads the Risk Analysis group at CSL (the UK's Central Science Laboratory which is part of the UK Government Department for Environment, Food and Rural Affairs).

Funded by the UK Engineering and Physical Sciences Research Council from 2001-04 was research into adsorption of pollutants on contaminated land. The goal of the project was to understand what attributes of pollutant and soil drive variability in adsorption so that these attributes can be incorporated into models used for risk assessment in relation to contaminated sites. More details can be found from the project description page.

Until 2001, my main area of research was methodology for modelling, calibration and forecasting for complex systems which have simulators which are used to understand their behaviour. Examples of such systems are oil reservoirs, gas and oil pipelines and areas of land contaminated by pollutants. What these and many other systems have in common is that the best way to understand the future behaviour of the system is to use a simulator, a computer implementation of a mathematical model for the system. The problem is that as part of the construction of a simulator, values have to provided for a great many quantities which describe physical, chemical, geological and other aspects of the system, quantities about which there is often a great deal of uncertainty. There is usually some data related to these quantities; for example one might have laboratory measurements of aspects of soil samples or the results of seismic or other sub-surface imaging. There may also be relevant expert opinion. Finally there will often be some historical data on the behaviour of the system which may be used to help calibrate the simulator by comparing it to the results of simulator runs for the historical period. For more detail, see the web page for our research group which has worked since 1993 on Bayesian methods for large physical systems with particular focus on methodology for production forecasting and history matching of hydrocarbon reservoirs.

I have a general interest in spatial probability and statistics which originated in post-doctoral research on the use of satellite imagery in mineral exploration.  With Allan Seheult, a DTI LINK funded project studied various aspects of a class of spatial Markov chains proposed by Qian and Titterington (JRSS B, 1990). On a consultancy basis, we have also produced a short course on geostatistics.

Like most statisticians, I try to provide statistical help and advice to researchers from other disciplines. Publications have arisen from work with Richard Hall (University of Leeds) on problems arising in hip-joint replacement and with Brian Diffey (Regional Medical Physics Unit) on some dermatological studies.

Recent Publications

• Bayesian Bayesian forecasting for complex systems using computer simulators (with Michael Goldstein, Jonty Rougier and Allan Seheult) (2001), Journal of the American Statistical Association, 96, pp 717-729.
• Constructing partial prior specifications for models of complex physical systems (with discussion) (with Michael Goldstein, Allan Seheult and James Smith) (1998),  The Statististician, 47, pp 37-53. Final draft available as an A4 Postscript file.
• Discrepancy between penetration depths derived from radiographic and direct measurement of acetabular components (with Richard Hall, P.Siney, A.Unsworth and B.M.Wroblewski) (1998), Proc. Instn Mech. Engrs, Part H, 212, pp 57-64.
• Posters about An elicitation tool for prior beliefs about complex systems (with Michael Goldstein), presented at the Practical Bayesian Statistics 4 meeting held at Nottingham University, July 1997.  Color Postscript files are available with the posters laid out on A3 paper, A3 pages split into 2 A4 pages, or as A3 pages shrunk to fit A4 pages.
• A prospective longitudinal study of the outdoor behaviour and symptoms of photosensitive patients (with Brian Diffey) (1997), British Journal of Dermatology, 137, pp 391-394.
• Pressure matching for hydrocarbon reservoirs: a case study in the use of Bayes linear strategies for large computer experiments (with discussion) (with Michael Goldstein, Allan Seheult and James Smith) (1997), in Gatsonis et al. (eds), Case Studies in Bayesian Statistics, Volume III, Springer, pp 37-93.
• Bayes linear strategies for matching hydrocarbon reservoir history (with Michael Goldstein, Allan Seheult and James Smith) (1996), in Berger et al. (eds), Bayesian Statistics 5, Oxford, pp 69-95.
• Pseudo-likelihood estimation for a class of spatial Markov chains (with Allan Seheult) (1996), Statistics and Computing, 6, pp 303-311.

Opportunities for Supervised Research

Other than the main research areas mentioned above, the list below gives some areas of research in statistics in which I would be particularly interested to supervise research students. If any of these are of interest to you, please contact me for more information.

You may also wish to read some more general information about postgraduate study in statistics at Durham

• Risk assessment for pesticides: calculations for human exposure n and/or ecological risk, use of variability factors, species sensitivity distributions.
• Elicitation of beliefs : computing environment, methods for complex systems, required complexity
• Modelling beliefs about functions: interesting function decompositions, deforming regular structures
• Image processing and stochastic geometry for assessment of severity of psoriasis
• Visual representation and manipulation of data analyses
• Bayes linear methodology for non-linear problems - image processing, spectrometer calibration
• Spatial Markov chains
• Dynamic linear models: non-Gaussian problems, identifiability, estimation and elicitation of covariance structure

Current Research Students

• Majid Sarmad is working on methodology and software for exploratory analysis of variance in the Tukey tradition.
• Jennifer Richardson is working on Bayesian methods for statistical modelling of infectious diseases
• I am also second supervisor for Vicky Montgomery and Rebecca O'Neil