Journal of Official Statistics, Vol.19, No.3, 2003. pp. 237–252
A Post-stratified Raking-ratio Estimator Linking National and State Survey Data for Estimating Drug Use
Trent D. Buskirk and Jane L. Meza
Abstract:Estimation of statewide and county-specific drug use can be improved by combining national and state survey data. Obtaining county estimates of drug use from state-level date is difficult due to the rarity of drug use and the small sample sizes within counties. In this article, we propose using post-stratified raking-ratio estimators to link state data with national surveys stratified by state. We also propose a raking estimator that links state census data to national survey data using auxiliary varibles associated with drug use. Specifically, the adjusted raking estimator produces estimates of drug use by region and age-class to be used in a small-area model that will produce empirical Bayes estimates of county drug use. The methods are presented using data from the Nebraska Adult Household Survey and the National Household Survey on Drug abuse. Two choices of small-area models are discussed.
Keywords:Post-stratified estimator; sampling weights; raking algorithm; small-area models; empirical Bayes estimates; borrow strength
Copyright © Statistics Sweden 1996-2018. Open AccessISSN 0282-423XCreated and Maintained by OKS Group