MRC-Biostatistics Seminar Series Spring Term 2006 10th January 2006 Title: Analyses of genetic association studies via haplotype clustering Professor David Balding Centre for Biostatistics Imperial College London Abstract: Approaches to the analysis of genetic association studies that go beyond using one marker at a time are often based on the notion of "haplotype", which can be thought of as a group of markers that are treated as a unit for analysis. There is support for this approach in the discovery over recent years that much of the genome has a block-like structure, with strong statistical dependence between markers within blocks, and little between blocks. However the "block" model of the human genome gives an imperfect description of reality, and current haplotype-based analyses have important limitations: haplotypes are not directly observed, but must be inferred from genotype data; it is difficult to model the relationships between haplotypes that are similar and so may have recent shared ancestry; and it is also difficult to accommodate rare haplotypes. I will discuss an approach to the analysis of genetic association studies that is based on a stochastic search for case-rich clusters of haplotypes. The talk will cover material that has just appeared in Genetic Epidemiology (http://www3.interscience.wiley.com/cgi-bin/jissue/107560885) plus some more recent developments. It is joint work with Ed Waldron (Imperial) and John Whittaker (LSH). 7th February 2006 Title: A general framework for parametrisation of hierarchical models Dr Omiros Papaspiliopoulos Mathematics and Statistics Department Lancaster University Abstract: In this talk we describe non-centring methodology for use in parametrisation of broad classes of hierarchical modelling, with a view to the construction of effective MCMC algorithms for exploring posterior distributions from these models. We give a clear qualitative understanding as to when non-centring works, and describe some theory on the convergence time complexity of Gibbs samplers using centred and non-centred parameterisations. We give general recipes for the construction of non-centred parameterisations, including an auxiliary variable technique called the state-space expansion technique. We also describe partially non-centred methods, and demonstrate their use in constructing robust Gibbs sampler algorithms whose convergence properties are not overly sensitive to the data. (The talk is primarily based on a paper with the same title to appear in the Statistical Science. It is joint work with G.O. Roberts and M. Skold). 7th March 2006 Title: Estimation of Occupancy Probabilities and Other Marginal Features in Multi-State Models Professor Jerry Lawless Department of Statistics and Actuarial Science University of Waterloo 14th March 2006 Title: Using the log-normal distribution to model bivariate survival Professor Gillian Raab Professor of Applied Statistics School of Community Health Napier University Abstract: With the current trend for modelling survival data as a counting process and summarising results in terms of hazard ratios use of the log-normal survival model has become unfashionable. I will illustrate its use for the bivariate survival problems with negative correlations between event times and argue that it may deserve to be used more often for survival problems. **************************************************************** The seminars starts at 2:30pm in the Large Seminar, Room, 1st Floor Institute of Public Health, University Forvie Site, Robinson Way, Cambridge. Tea and coffee are served afterwards. All are welcome to attend. (For a full list of seminars and abstracts please see www.mrc-bsu.cam.ac.uk/BSUsite/Education/seminars.shtml)