A1141
Title: Parsimonious ultrametric Manly mixture models
Authors: Paul McNicholas - McMaster University (Canada) [presenting]
Alexa Sochaniwsky - McMaster University (Canada)
Abstract: A family of parsimonious ultrametric mixture models with the Manly transformation is developed for clustering high-dimensional and asymmetric data. Advances in Gaussian mixture modelling sufficiently handle high-dimensional data but struggle with the presence of cluster skewness. Also, while these advances reduce the number of free parameters, they often provide limited insight into the structure and interpretation of the clusters. To address this shortcoming, the extended ultrametric covariance structure and the Manly transformation are used, resulting in the parsimonious ultrametric Manly mixture model family. Model selection proves challenging, and a two-step model selection procedure is proposed. Simulation studies and real data analyses are used for illustration.
