Multivariate cluster weighted models using skewed distributions

Much work has been done in the area of cluster weighted model (CWM), which extends finite mixture regression to include modelling covariates. Although many types distributions have considered for both response(s) and covariates, our knowledge skewed not yet this paradigm. Herein, a family 24 novel CWMs is allows responses covariates be modelled using one four (the generalized hyberbolic three its special cases, i.e., skew-t, variance-gamma normal-inverse Gaussian distributions) or normal distribution. Parameter estimation performed expectation-maximization algorithm simulated real data are used illustration.