Pipeline network and oil reservoir modelling
This work concerns Bayesian approaches to very complicated stochastic systems, for example the modelling of leak detection for pipeline networks and history matching for oil reservoirs. The consultancy, which is collaborative with Energy Scitech Ltd., has arisen out of an EPSRC research grant, entitled Bayes linear forecasting and decision-making for large-scale systems in the petroleum industry, and aims at implementing aspects of that research commercially.
Improving manufacturing performance
Manufacturing companies need continually to improve performance to keep or achieve a competitive edge. In particular, companies need to make their products better (higher quality, fewer returns) and cheaper (faster production, lower waste). With a number of companies we have investigated production lines using experimental design methodology. For one such company we reduced wastage from 18% to 3% and simultaneously improved throughput by 30%.
Bayesian graphical models for software testing
This work concerns a major research study in collaboration with British Telecommunications Plc and statisticians and computer scientists at the University of Durham. The basic aim of the work is to arrive at an improved statistical methodology for treating software testing, exploiting the expertise of BT staff, in order to arrive at even more reliable software, more quickly and more cheaply than current methods allow. The success of this project has led to an associated research grant, funded by EPSRC, entitled High reliability software testing for complex software using Bayesian graphical modelling and program comprehension.
A local company has 3500 products, of which around 600 per day can be manufactured. Therefore the warehouse must be large enough to hold sufficient stock of each kind of product so that orders can be met, from stock and newly scheduled production, with small probability of a stockout. We model the demand for products, taking into account possible production schedules, and advise on stock levels for each product, assuming a range of stockout probabilities to be decided by managers.
Design and analysis of surveys for social policy decisions
This work is collaborative with colleagues in the Centre for Applied Social Studies, University of Durham, and social scientists elsewhere. We provide the statistical expertise underpinning several different projects relating to social policy decisions, for example regarding the care of people with mental health or learning disability problems. We help to design the studies, provide advice on the quality of the measurement instruments employed, and analyse and explore the resulting surveys.
Modelling of sub-arachnoid haemorrhage
This work concerns statisticians working with experienced physicians in Accident and Emergency departments in the main North Eastern hospitals to improve our understanding of sub-arachnoid haemorrhage (SAH). This is a rare but often fatal disease involving the bursting of blood vessels at the periphery of the brain. It is hypothesized that ``warning leaks'' (minor haemorrhages) can precede major (usually fatal) haemorrhages. Properly diagnosed warning leaks lead to treatment which can prevent a further major haemorrhage. However, there is a suspicion that such warning leaks, for which the major symptom is severe headache, are often misdiagnosed for a variety of reasons. The work is thus aimed at exploring whether it is possible to identify particular combinations of patient signs and symptoms that indicate SAH with high probability. The approach we take is to construct Bayesian graphical models relating patient characteristics and symptoms and other diagnostic and pathognomonic factors to the possible underlying diseases.
Bird survival modelling
Birds are captured, ringed and released back into the wild. Over periods of several years, many observers set out to record resightings of individual birds. Of interest is the survival distribution for birds, whether or not this changes over time, and whether or not birds' survival is related to climatic changes. The approach we take is based on Bayesian graphical modelling using the software.
Bayesian forecasting of financial series
Where a company has to provide forecasts of financial series, such as commodity prices, the forecast is often based on a number of factors, typically measurements (data) and expert judgements. For one such project we are working with the company to combine these factors, using Bayesian statistics, to provide forecasts which take better account of the variety of underlying factors and their compatibility, and in particular to understand the uncertainties attached to them and to calibrate to new data.
The statistics of child smoking
This concerns analysis of a large survey exploring child smoking behaviour. Nearly 14,000 children in North-Eastern England gave confidential details of their smoking habits, likes and dislikes, familial smoking patterns, and attitudes to smoking by friends, relations, and teachers. A book presenting the results is shortly to appear.
Training in statistical methods for workers in the automotive industry, taking account of the background of the company personnel and their working environment. Some training focuses on underpinning the concepts of six-sigma, a popular approach to improving product quality and manufacturing performance.