Uncertain stochastic ridge estimation in partially linear regression models with elliptically distributed errors

In fitting a regression model to survey data, using additional information or prior knowledge, stochastic uncertainty occurs in specifying linear programming due to economic and financial studies. These stochastic constraints, definitely cause some changes in the classic estimators and their efficiencies. In this paper, stochastic shrinkage estimators and their positive parts are defined in the partially linear regression models when the explanatory variables are multicollinear. Also, it is assumed that the errors are dependent and follow the elliptically contoured distribution. The exact risk expressions are derived to determine the relative dominance properties of the proposed estimators. We used generalized cross validation (GCV) criterion for selecting the bandwidth of the kernel smoother and optimal shrinkage parameter. Finally, the Monte-Carlo simulation studies and an application to real world data set are illustrated to support our theoretical findings.