Professor John Haslett

Biography

Apart from sabbatical visits to Seattle, Canberra, Sydney and Massey Univ in NZ, I have been in TCD since graduate school in Birmingham. My current interests are focussed on spatio-temporal statistical models and 'related areas'. Applications to wind energy and to palaeo-climate reconstruction have led to modelling ideas which have appeared in papers that have been 'read' the Royal Statistical Society (1989 and 2006, in press). The methodologies have included long memory time series and Markov chain Monte Carlo (MCMC). In particular we have developed fast MCMC methods for leave-k-out cross-validation in 'inverse' models. Other past applications have included digital satellite imagery as used in mineral exploration. Current (Nov 2005) research proposals include applying the Bayesian methods used in our palaeo-climate work to (a) the spread of the Neolithic culture through Europe (with colleagues in Genetics and Archeology) and (b) modern official statistics (with the Central Statistics Office). A related area is that of General Least Squares (GLS), the methodology under-pinning 'geo-statistics', a widely modelling framework used in spatial data analysis. GLS methods underpin many modern and classical statistical procedures. Thus leave-k-out deletion diagnostics, developed for geo-statistics, have very wide application to areas including random coefficient models.