CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: Gaussian parsimonious clustering models with covariates Authors:  Keefe Murphy - University College Dublin (Ireland) [presenting]
Thomas Brendan Murphy - University College Dublin (Ireland)
Abstract: Model-based clustering methods are considered which account for external information available in the presence of covariates by introducing the MoEClust family of models, and a related software implementation. These finite mixture models allow the distribution of the latent cluster membership variable and/or the distribution of the response variables to depend on fixed covariates, under a range of parsimonious eigen-decomposition parameterisations of the component covariance matrices. Thus, the following equivalent aims are addressed: including covariates in Gaussian parsimonious clustering models and incorporating parsimonious covariance structures into the Gaussian mixture of experts framework. The MoEClust models demonstrate significant improvement from both perspectives in applications to data with multivariate responses and covariates of mixed type and provide richer insight into the type of observation which characterises each cluster.