Mixtures of conditional mean- and covariance-structure models

Models and parameters of finite mixtures of multivariate normal densities c onditional on regressor variables are specified and estimated. We consider mixtures of multivariate normals where the expected Value for each componen t depends on possibly nonnormal regressor variables. The expected values an d covariance matrices of the mixture components are parameterized using con ditional mean- and covariance-structures. We discuss the construction of th e likelihood function, estimation of the mixture model with regressors usin g three different EM algorithms, estimation of the asymptotic covariance ma trix of parameters and testing for the number of mixture components. In add ition to simulation studies, data on food preferences are analyzed.