Gravar-mail: Regression mixture models: Does modeling the covariance between independent variables and latent classes improve the results?