Title: Random regression coefficients model for small area estimation
Authors: Gauri Datta - University of Georgia and US Census Bureau (United States) [presenting]
Hee Cheol Chung - University of Georgia (United States)
Jerry Maples - US Census Bureau (United States)
Abstract: Small area estimation methodology has been found to be an indispensable tool to reliably estimate various government statistics such as income, employment, poverty status, health care availability, disease prevalence, etc. for various segments of a population. Sample surveys have been effectively used to provide suitable statistics not only for the population as a whole targeted by a survey but also for a variety of subpopulations, often called domains or areas. Domains may be geographical areas such as states, districts, or socio-demographic groups or other subpopulations. Often samples collected from some domains are not large to produce on their own accurate statistics for those domains. These domains are considered ``small areas" which need alternative estimates with better accuracy. Small area estimation methodology benefited immensely from Stein's shrinkage estimation by borrowing strength from the direct estimates of the other areas and appropriate auxiliary variables available for all the areas. A part of the variability of the population small area means is explained through a suitable regression model based on auxiliary variables. We present a noninformative Bayesian analysis of random regression coefficients model to produce reliable point estimates of population means. Our method generalizes the standard Fay-Herriot model.