New Bayesian Estimation for Single Index Model

Abstract

Generalization of the semi-parametric single index model to be more flexible  than the general linear model by allowing non-linear relationships between the index function  and the response variable. In this paper,  new estimation and variable selection method through Bayesian approach is proposed.  We have  construct new hierarchical model based on  the representation  of scale mixture of normal distribution mixing Rayleigh density for the  double exponential prior density of the parameters vector. Two simulation examples and real data are considered to evaluation our proposed method compare to some existing methods and we get some results.