Parsimonious mixture of mean-mixture of normal distributions with missing data

Clustering multivariate data based on mixture distributions is a usual method to characterize groups and label data sets. Mixture models have recently been received considerable attention to accommodate asymmetric and missing data via exploiting skewed and heavy-tailed distributions. In this paper, a mixture of multivariate mean-mixture of normal distributions is considered for handling missing data. The EM-type algorithms are carried out to determine maximum likelihood of parameters estimations. We analyzed the real data sets and conducted simulation studies to demonstrate the superiority of the proposed methodology.

Keywords:

EM-type algorithms , Finite mixture model , MMN distribution , Missing Data , Skew distribution