Robust mixture regression modeling based on two-piece scale mixtures of normal distributions

The inference of mixture regression models (MRM) is traditionally based on the normal (symmetry) assumption of component errors and thus is sensitive to outliers or symmetric/asymmetric lightly/heavy-tailed errors. To deal with these problems, some new mixture regression models have been proposed recently. In this paper, a general class of robust mixture regression models is presented based on the two-piece scale mixtures of normal (TP-SMN) distributions. The proposed model is so flexible that can simultaneously accommodate asymmetry and heavy tails. The stochastic representation of the proposed model enables us to easily implement an EM-type algorithm to estimate the unknown parameters of the model based on a penalized likelihood. In addition, the performance of the considered estimators is illustrated using a simulation study and a real data example.