Comparing two formulations of skew distributions with special reference to model-based clustering
Adelchi Azzalini, Ryan P. Browne, Marc G. Genton, Paul D. McNicholas · arXiv (Cornell University) · 2014
Multivariate skew distributions continue to gain popularity as effective tools in data analysis for a wide variety of data types. They also feature in several modern computational statistics techniques, including model-based clustering approaches. Two well-known formulations of skew distributions have been used extensively within the model-based clustering literature over the past few years. We investigate the properties of these formulations and, in doing so, we refute various claims that have been made about these formulations of late. Our position as to the inaccuracy of these claims is supported by theoretical arguments as well as real data examples. 1