Responsive Designs for Rare Subpopulations Subject to Misclassification

• Activity Number: 287
• Type: Topic Contributed
• Date/Time: Tuesday, August 2, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Government Statistics
• Author(s): Randall Powers*+ and John L. Eltinge
• Companies: Bureau of Labor Statistics and Bureau of Labor Statistics
• Keywords: Efficient design ; Measurement error ; Point estimation ; Response error ; Subsampling ; Total statistical risk

Abstract

When working with data from surveys or administrative records, one often encounters special issues in the estimation of prevalence rates and means for rare subpopulations. Two notable issues arise when some data are subject to misclassification, and when one obtains limited data for some of the subpopulations of interest. This paper explores the extent to which one may adapt responsive-design methods to produce improved estimators of both prevalence rates and subpopulation means. In this paper we propose an adaptive design and methods for estimating prevalence rates and subpopulation means under this design. We then provide a detailed simulation study focused on evaluation of bias and mean squared error for the proposed estimators under specified sets of conditions.