Skip to main content

Seminar: Estimation for nearest neighbor imputed survey data

Jul 7, 2021 - 10:10 AM
to , -

Abstract:  Sample surveys usually have missing data and have multiple response variables. Nearest neighbor imputation is one procedure used to complete missing records. However, the direct nearest neighbor imputation estimator suffers from a bias that increases as the dimension of the covariate vector increases. In this paper, we construct a regression estimator for the population mean based on the nearest neighbor imputed dataset. We use the regression procedure to reduce the bias in the direct nearest neighbor estimator. Under the linear model, the proposed estimator is unbiased with a smaller mean square error than the direct nearest neighbor estimator. We give a replication variance estimator which does not require repeated imputation. A simulation study is conducted to compare the efficiencies of the proposed estimator and the direct nearest neighbor estimator for the population mean.