Two families of parsimonious mixture models are used for model-based clustering. They are based on the multivariate shifted exponential normal and the multivariate tail-inflated normal distributions, heavy tailed generalizations of the multivariate normal. Parsimony is achieved via the eigen-decomposition of the component scale matrices, as well as by imposing a constraint on the tailedness parameter. Two variants of the expectation-maximization algorithm are used for parameter estimation. Identifiability conditions are illustrated, and the advantages of our models with respect to other existing parsimonious heavy-tailed mixture models are commented. Our models are firstly tested via simulation studies, and then compared to some competing models in real data applications.

Clustering via new parsimonious mixtures of heavy tailed distributions

Tomarchio S. D.
;
Punzo A.
2021

Abstract

Two families of parsimonious mixture models are used for model-based clustering. They are based on the multivariate shifted exponential normal and the multivariate tail-inflated normal distributions, heavy tailed generalizations of the multivariate normal. Parsimony is achieved via the eigen-decomposition of the component scale matrices, as well as by imposing a constraint on the tailedness parameter. Two variants of the expectation-maximization algorithm are used for parameter estimation. Identifiability conditions are illustrated, and the advantages of our models with respect to other existing parsimonious heavy-tailed mixture models are commented. Our models are firstly tested via simulation studies, and then compared to some competing models in real data applications.
978-88-5518-340-6
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/20.500.11769/535018
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact