Trend Analysis for Complex Survey Data with Bayesian Approach

• Activity Number: 295 - SPEED: Big Data, Small Area Estimation, and Methodological Innovations Under Development, Part 1
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 8:30 AM to 10:20 AM
• Sponsor: Survey Research Methods Section
• Author(s): Yi Mu*
• Companies: Centers for Disease Control and Prevention
• Keywords: trend; complex survey; Bayesian; infectious disease

Abstract

Researchers often use complex sample survey methodology to obtain information about a large population by selecting and measuring a sample from that population. Analyzing the sample data may involve estimating the unknown subpopulation quantities (domains) that are aggregated over a stratified group. As a result, the population is further stratified, the estimated weights for each new stratum can be calculated and applied to the traditional survey software for secondary analysis. However, the traditional approach would not be able to account for the variances associated with the estimated domains. In this study, we proposed to use Bayesian approach: to build an error model with the ‘true’ domain value depending on the estimate and the standard error associated with it. This error model can be plugged into any regression model for further analysis. Both traditional and Bayesian approach would be applied and compared in the context of the trend analysis of Clostridioides difficile infection, a bacterium that causes diarrhea and colitis and estimated to infect almost half a million Americans each year.