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.
机构:
E China Normal Univ, Dept Stat, Shanghai 200062, Peoples R China
Fudan Univ, Sch Publ Hlth, Dept Biostat, Shanghai 200032, Peoples R ChinaE China Normal Univ, Dept Stat, Shanghai 200062, Peoples R China
Qin Guoyou
Zhu Zhongyi
论文数: 0引用数: 0
h-index: 0
机构:
Fudan Univ, Dept Stat, Shanghai 200433, Peoples R ChinaE China Normal Univ, Dept Stat, Shanghai 200062, Peoples R China