Mixture models have been popular in the literature of both statistics and social science. In this dissertation, we propose a new mixture model, namely, nonparametric finite mixture of regression models, which can be viewed as a natural extension of finite mixture of linear regression. In the...