Calibrated Bayesian Approach for Small Area Prevalence Estimation Using Survey Data with Replicate Weights

• Activity Number: 340 - SPEED: Applications of Advanced Statistical Techniques in Complex Survey Data Analysis: Small Area Estimation, Propensity Scores, Multilevel Models, and More
• Type: Contributed
• Date/Time: Tuesday, July 31, 2018 : 10:30 AM to 12:20 PM
• Sponsor: Survey Research Methods Section
• Author(s): Trung Ha* and Julia Soulakova
• Companies: University of Central Florida and University of Central Florida
• Keywords: balanced repeated replications; complex sampling; design-based estimation; model-based estimation; secondhand smoke; smoking rules at home

Abstract

National surveys commonly incorporate complex sampling, which requires use of the pre-specified replicate weights in data analysis. Small area estimation problems arise when one wants to perform estimation for geographical and non-geographical areas with small counts using the national survey data. We propose a hybrid approach for small area prevalence estimation that incorporates design-based estimation and model-based estimation using Bayesian Logistic Linear Mixed model. The proposed method belongs to the class of Little's Calibrated Bayesian methods. The approach is appropriate even when the area proportions are close to zero or one. We illustrate how the hybrid method can be used for estimating proportions of smoke-free homes among single-family households for 10 parental racial/ethnic groups in the U.S. using the 2014-15 Tobacco Use Supplement to the Current Population Survey (CPS) data. The hybrid approach offered more informative interval estimators for several racial/ethnic groups in comparison to the design-based approach. Generalizations to many other complex surveys are expected to be straight-forward.