The Mixed Estimator of Truncated Spline and Local Linear in Multivariable Nonparametric Regression

In the multivariable nonparametric regression method, modeling is generally done by using one type of estimator for all predictor variables. This is due to the assumption that each predictor is considered to have the same data pattern so that the analysis only uses one type of estimator for all predictor variables. However, in reality, there are often cases where each predictor variable has a different pattern. As a result, the regression model estimation becomes less precise and tends to produce large errors. To overcome this problem, the researcher developed a mixed estimator involving two types of estimators in the model. This article provides an overview of the mixed estimator method in multivariable nonparametric regression. The method using a mixed estimator of Truncated Spline and Local Linear. The mixed estimator is obtained by solving the two-stage estimation, consisting of weighted least square (WLS) and least square (LS) optimization. This estimator is able to accommodate multivariable nonparametric regression models in various real cases that involve more than one predictor, in which there are some predictors follow spline characteristics and other predictors follow local linear characteristics.