Keywords: nonparametric regression, machine learning, survey asymptotics
Model-assisted estimation provides a framework for incorporating auxiliary information in survey estimation and inference. Originally developed for linear models, the approach has been shown to be applicable to many other types of relationships between variables. In this talk, we discuss a general "recipe" for deriving model-assisted estimators and for the associated design-based asymptotic analysis. The recipe allows for a very broad class of prediction methods, and we give a review of some of the newer methods involving nonparametric and machine learning techniques.