Title: Mixture modelling of high-dimensional data
Authors: Geoffrey McLachlan - University of Queensland (Australia) [presenting]
Abstract: Some aspects of the use of finite mixture distributions in modelling high-dimensional data are considered. Attention is focussed on mixtures with multivariate normal and t-distributions and some skew variants for the component distributions. Dimension reduction is undertaken via factor models which allow for skew distributions in addition to white noise for the factor distributions. Consideration is also given to dimension reduction via clustering of the variables. Applications are given involving the analyses via mixture modelling of some real data sets in the biomedical and health sciences.