Title: Continuous mixture of normal distributions for cluster analyses
Authors: Sharon Lee - University of Adelaide (Australia) [presenting]
Abstract: The continuous mixture of normal distributions generalizes the normal distribution by scaling its mean, variance, or both with a (continuous) random variable, thereby allowing more flexible distributional shapes such as heavy tailedness and skewness. This renders them useful for modelling non-normal data. We present a selective overview of the continuous mixture of normal distributions suitable for model-based clustering. In particular, we consider the families of location mixture, scale mixture, and a location-scale mixture of normal distributions. We discuss their basic properties, important special/limiting cases, and methods for parameter estimation.