Estimation of Semiparametric Regression Function in Presence of Measurement Error

Abstract

In recent years, the attention of researchers has increased on semi-parametric regression models, because it is possible to integrate the parametric and non-parametric regression models in one and then form a regression model that has the potential to deal with the case of dimensionality in non-parametric models. This occurs through the increasing of explanatory variables. involved in the analysis and then decreasing the accuracy of the estimation. As well as the privilege of this type of model with flexibility in the application field compared to the parametric models which comply with certain conditions such as knowledge of the distribution of errors or the parametric models may not represent the phenomenon properly studied. In this paper, we will show semi-parametric methods in estimation of regression function in the presence of measurement error, and this methods are Simex method and instrument variable method and Quassi-likelihood method and will compare between these methods by using ( MASE ) criterion. A simulation had been used to study the empirical behavior for the semi-parametric models, with different sample sizes and variances. The results showed that the instrument variable is better than simex and Quassi-likelihood methods at different sample sizes and variances that been used.