Finite Mixtures of Individual-Level Models for Infectious Diseases

• Activity Number: 592
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Rob Deardon*+ and Lorna Deeth
• Companies: University of Guelph and University of Guelph
• Keywords: infectious diseases ; MCMC ; finite mixture models ; individual-level models ; Bayesian inference

Abstract

A class of models, known as individual-level models (ILMs), has been used to describe the spatiotemporal spread of infectious diseases. The current approach to ILMs does not account for hidden population heterogeneity; instead, it assumes a homogeneous population and consequently, model parameters common to the entire population are used. However, if a latent grouping structure exists within the population, this assumption may be inaccurate, as the parameters driving disease spread within one subgroup may be significantly different from those in another. Finite mixture models are a way to circumvent the problem of population heterogeneity, by incorporating the group structure into the modeling procedure. We therefore propose a finite mixture of ILMs model that is comprised of component ILMs, and consider the use of this model with both real and simulated data.