Linear regression models using finite mixtures of skew heavy-tailed distributions

In this paper, we propose a regression model based on the assumption that the error term follows a mixture of normal distributions. Specifically, we consider a finite scale mixture of skew-normal distributions, a rich family that contains the skew-normal, skewt,
skew-slash and skew-contaminated normal distributions as members. This model allows us to describe data with high flexibility, simultaneously accommodating multimodality, skewness and heavy tails. We develop a simple EM-type algorithm to perform maximum
likelihood inference of the parameters of the proposed model with closed-form expressions for both E- and M-steps. Furthermore, the observed information matrix is derived analytically to account for the corresponding standard errors and a bootstrap procedure is implemented to test the number of components in the mixture. The practical utility of the new model is illustrated with a real dataset and several simulation studies. The proposed algorithm and methods are implemented in an R package named FMsmsnReg.