ABSTRACT
Many variables of interest in medical and health sciences can be treated as continuous data. When clusters in the data are essentially elliptical or any non-normal features in the data are attributed to some underlying group structure, then a mixture of elliptically symmetric component distributions could be used to explore the group structure of the data. This chapter describes advanced developments of the normal mixture model within the mixture modelling literature. This includes the mixture of skew normal or skew-t distributions, the mixture of normal inverse Gaussian distributions, the mixture of log-concave densities, the mixture of linear mixed-effects models, the mixture of sparse regression models, and the mixture of matrix normal distributions. Developments to normal mixture models have been made for the purposes of handling matrix data including spatial multivariate data, multivariate time series, and longitudinal vector measurements in a wide variety of experimental settings.
