Mixtures of Factor Models for Multivariate Disease Rates

Abstract

A range of different approaches have been suggested for the multivariate modelling of the geographical distribution of different but potentially related diseases. We suggest an addition to these methods which incorporates a discrete mixture of latent factors, as opposed to using CAR or MCAR random effect formulations. Our proposal provides for a potentially richer range of dependency structures than those encompassed in previously used models in that it is capable of representing an enhanced range of correlation structures between diseases at the same time as implicitly allowing for less restrictive spatial correlation structures between geographical units. We illustrate results of using the model on data taken from cancer registries on four carcinomas in some 300 UK geographical areas.