Imputation Procedures in Surveys Using Nonparametric and Machine Learning Methods: An Empirical Comparison

Abstract Nonparametric and machine learning methods are flexible for obtaining accurate predictions. Nowadays, data sets with a large number of predictors complex structures fairly common. In the presence item nonresponse, nonparametric procedures may thus provide useful alternative to traditional imputation deriving set imputed values used next estimation study parameters defined as solution population estimating equation. this paper, we conduct an extensive empirical investigation that compares in terms bias efficiency wide variety settings, including high-dimensional sets. The results suggest perform very well efficiency.