A new method for solving the problem of the mean estimation when the underlying regression function is discontinuous

Abstract

In the context of finite population survey sampling, we propose a new model-based mean estimator, when the function that links the variables is discontinuous. The available estimators of the mean based on nonparametric regression are derived under the assumption that the regression function is continuous. We propose a new approach to adjust for the effect of discontinuity on regression estimation of the mean. The performance of the proposed estimator is analysed through a simulation study because the theoretic study of asymptotics is not possible. In the literature, the new estimator requires more computational cost than others, but the simulation experiments indicate that the proposed method has higher efficiency than other traditional parametric and nonparametric regression methods.

Keywords:

• discontinuity
• jump point
• local polynomial kernel regression
• model-assisted approach
• Horvitz–Thompson estimator

2000 AMS Subject Classification :

• 62D05
• 62G08
• 68U20