ABSTRACT
This chapter introduces mixtures of linear mixed models with two sets of random effects and describes how the models provide a flexible approach for clustering multivariate continuous data collected using various experimental settings, including cross-sectional repeated-measure data, time-course data, and multilevel longitudinal data. Mixture model-based approaches have been widely used in the cluster analysis of gene expression data given that mixture models provide a sound mathematical framework for clustering. Time-course experiments with microarrays are often performed to study dynamic biological systems and genetic regulatory networks that model how genes influence each other in cell-level development of organisms. It is well known that the information encoded in DNA leads to the expression of certain phenotypes in the organism. For clustering longitudinal data in the form of growth trajectories, growth mixture models and mixture latent growth models developed by J. K. Vermunt and Vermunt have been adopted to identify different clusters of trajectory patterns and predictors of membership in these classes.
