A Comparison of Design-Based and Calibrated Bayes Estimates Using Data from a Health Survey
Meena Khare
National Center for Health Statistics, CDC
Alena Maze
National Center for Health Statistics
Hee-Choon Shin
National Center for Health Statistics
Bayesian methods have been gaining popularity as an alternative to the traditional design-based methods for estimation from complex surveys. In this paper, we apply calibrated Bayes methods to estimate vaccination rates from the National Immunization Survey (NIS). NIS is a large telephone survey, which has been continuously conducted to monitor childhood vaccination coverage among U.S. children aged 19-35 months since 1994 (www.cdc.gov/nchs/nis.htm). Official design-based vaccination coverage rates at the national, state, and selected urban area levels estimates using data from the NIS are available www.cdc.gov/vaccines/stats-surv/nis/default.htm #nis. Data from the recent NIS public-use files are used to compute and compare the Bayesian estimates with the design-based estimates. We also compare subdomain estimates based on the two methods by selected demographic characteristics.