Imputation Based On Local Polynomial Regression for Nonmonotone Nonrespondents in Longitudinal Surveys

Abstract

In this study, the problem of nonrespondents in longitudinal survey’s data is considered. The study focuses on the imputation for the longitudinal survey data which often has nonignorable nonrespondents. Local linear regression is used to impute the missing values of and then the estimation of the time-dependent finite populations means. The estimation of the time dependent means was based on the assumption that the nonresponse mechanism is last past value dependent. The asymptotic unbiasedness and consistency of the proposed estimator are investigated. The imputation for the nonmonotone nonrespondents is done multiple times through simulation and the simulation study is carried out to asses the best performing estimator of the time-dependent finite populations means. Comparisons between different parametric and nonparametric estimators are performed based on the bootstrap standard deviation, mean square error and percentage relative bias. The simulation results show that local linear regression estimator yields good properties.