Model-assisted Estimation For Surveys Using Non-Parametric Methods

With the advent of the era of big data,data has received more attention than ever before.Data collection and analysis have played an important role in daily life,business management,and macroeconomic regulation of the national economy.The goal of the paper is restricted to sampling survey which is the most important mean of collecting data at present.The paper explores how to utilize auxiliary information through the nonparametric methods in the phase of sampling estimation to estimate the total of the population.Specifically,this paper does a research of nonparametric methods in model-assisted estimation on the following aspects.Firstly,by summarizing the existing model-assisted estimation theories,this paper built a unified framework of model-assisted estimation through analyze the connection of design-based HT estimator and the generalized difference estimator.Secondly,based on reviewing the model-assisted non-parametric methods and taking the model calibration method into account,a new class of nonparametric model-assisted estimator is proposed to make up for the deficiency of the traditional estimators.Under standard design conditions,the estimators are design consistent and asymptotically normal,and they admit consistent variance estimation using familiar generalized difference estimate.Simulation experiments show that the estimators are more efficient than parametric regression estimators when the parametric model is incorrectly specified,while being approximately as efficient when the parametric specification is correct.Thirdly,the sampling survey practice and study in our country are lagging behind the advanced level abroad,the paper not only could provide a theoretical basis for related people,but also make a contribution to the application and the development of model-assisted approach.