ESTIMATION IN A PARTIAL REGRESSION MODEL FOR CONTAMINATED DATA USING THE ROBUST PARTIAL RESIDUAL ESTIMATION TECHNIQUE

The partial regression model is a type of semi-parametric regression model that takes the form of a simplified semiparametric regression model. The partial regression model is more flexible than the parametric linear models because it combines parametric and non-parametric components in which the response variable has a linear relationship with some explanatory variables and a non-linear relationship with others. In this study, we examine two estimation methods, the robust partial residuals estimation method, and the non-robust estimation by Nadaraya-Watson smoothing and Speakman estimation, and compare them using mean squared error (MSE) as a comparison criterion for simulation experiments at various sample sizes, contaminated rates and a set of default values for the.. By comparing the two estimators, it became evident that the robust partial residuals are more efficient than the non-robust estimator since, it has the lowest mean square error (MSE) and is real parameter values approximate the default parameter values.