A solution in small area estimation problems

TL;DR

This paper introduces a novel small area estimation method that leverages auxiliary information about population unit sizes, incorporating variability and domain positioning to improve estimates, especially with small or zero sample sizes.

Contribution

It presents a new estimation approach using auxiliary size information and a regression-based adjustment to reduce bias and improve accuracy in small area estimation.

Findings

Method improves estimation accuracy in small domains.

Simulation confirms efficiency over existing methods.

Incorporates variability and domain position into estimators.

Abstract

We present a new method in problems where estimates are needed for finite population domains with small or even zero sample sizes. In contrast to known estimation methods, an auxiliary information is used to model sizes of population units instead of a direct prediction of their values of interest. In particular, via an additional characterization of regression models, we incorporate a scatter and variabilities of the units sizes into an estimator, and then it uses an information of the whole sample by taking into an account a location of the estimation domain inside the population. To reduce an impact of the introduced domain total estimator bias to the mean square error, we construct also a regression type version of the estimator. An efficiency of the method proposed is shown in a simulation study.