School of Mathematical Sciences

Research menu

Collaborative research

Collaborative Research between Statisticians and People in Other Disciplines

As statisticians, we are very happy to get involved in genuine research collaborations with people in other disciplines. We prefer to have ongoing collaborations over a period of time, so that we can obtain some understanding of the problems, vocabulary and techniques of the other discipline. In particular, we like to be involved in the design of the study as well as in the analysis of data from it.

Below are some examples of such collaborations.


R. A. Bailey and S. G. Gilmour (once at Queen Mary, now at Southampton) are collaborating with D. Perkins, J. Reiss (once at Queen Mary, now at Roehampton) and G. Woodward in the School of Biological and Chemical Sciences. The aim is find out about the effects of biodiversity on certain biological processes, such as the consumption of leaf litter in freshwater streams. Experiments have been devised; a novel family of linear models has been fitted; and guidance is being developed to allow the biologists to fit these non-standard models using available software. The findings from one of these experiments challenge the received wisdom about the effects of biodiversity.

Bumblebee personality

H. Großmann is working with H. Muller and L. Chittka in the School of Biological and Chemical Sciences to investigate the behaviour of bumblebees. One goal of the project is to investigate whether bees behave consistently, and several experiments to answer this question have been performed. As part of the statistical input, a mixed model which incorporates several fixed and random effects likely to influence bee behaviour and also takes into account the repeated measurements nature of the data has been developed and used to analyse the data.

Diabetes therapy

L. I. Pettit is collaborating with K. D. S. Young (University of Surrey), A. Onar (St Jude's children's research hospital, Memphis, USA) and R. Holman and S. Paul (Diabetes Trials Unit, University of Oxford). The aim is to use a Bayesian analysis of a degradation model to help predict how quickly patients diagnosed with type-2 diabetes will have to change initial therapy (diet or drugs) based on their baseline characteristics.

Pharmaceutical industry

B. Bogacka collaborates with statisticians and experimenters in Pfizer and in Novartis on some design issues in pre-clinical studies.

Uncertainty of a contaminant concentration

D. S. Coad has worked with M. H. Ramsey in the School of Life Sciences at the University of Sussex and J. A. Lyn at the Food Standards Agency on two-stage nested designs with random effects for assessing uncertainty of a contaminant concentration. The aim is to develop approximate confidence intervals for the variance components for unbalanced designs when the data are normally distributed. Since, in practice, there are often outliers and the data cannot be assumed to be normal, robust statistics can be used.