My primary research interest is computational statistics, with a particular focus on Markov chain Monte Carlo (MCMC). I first became interested in MCMC during the final year of my undergraduate degree, and subsequently developed that interest throughout my Ph.D. My current position at Imperial is allowing me to build on this experience to develop new algorithms for genetic and epidemiological applications.

My postdoc position at Imperial, under Prof. Sylvia Richardson is focussed on trying to tease out combinations of epidemiological variables that are significant in the development of lung cancer. To achieve this I am working on methodological developments for modelling statistical interactions. These developments are broadly falling into two categories: profile regression and inclusion of interactions in variable selection methods for regression involving large p small n problems.

Profile regression is a relatively new Bayesian technique that clusters observations according to their covariate profile, and maps clusters to outcomes. My initial contributions in this direction are two-fold: firstly in order to apply the method to large and more diverse applications I am re-writing existing software in C++; and secondly I am exploring ideas to make the output of the method more interpretable, in particular when comparing the output from applying the method to two distinct data sets.

My work on Bayesian variable selection has been specifically been on the Evolutionary stochastic search algorithm, an Evolutionary Monte Carlo based approach for single and multiple response linear models. I am working with my collaborators on extending the algorithm to be applicable to logistic regression and regression with interaction terms. I have been involved in developing the C++ software for this algorithm am also working on using GPU programming techniques to improve performance of the algorithm.

I studied for my Ph.D. in the Statistics Group at the University of Bristol, under the supervision of Prof. Peter Green. My thesis was on the subject of reversible jump MCMC, and specifically the development of an automatic sampler requiring minimal user input and no expert tuning. I also spent 6 months of my Ph.D. in the Biostatistics Department in the University of Wisconsin, Madison, working under Prof. Michael Newton. My thesis is available in the publications page.

Between completing my Ph.D. and returning to academia at Imperial I worked for 7 years in the Quantitative Team at Smartodds, primarily developing statistical models for football. For a nice paper describing a robust building block for such models see this paper, written by my good friend Stuart Coles, who continues to work at Smartodds. Our work extended this research in many directions, but remains proprietary to Smartodds and cannot be shared.