Estimation and inference for mixture of partially linear additive models

Abstract

In this paper, a semiparametric mixture of regression models is proposed, where the regression functions are partially linear additive while the mixing proportions and variances are unknown constant. The asymptotic normality of the SBK estimators and the model selection consistency of the proposed BIC based on B-spline estimators are established. Simulations and applications are presented to illustrate the performance of the proposed.