|
Activity Number:
|
107
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 9, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #301586 |
|
Title:
|
A Bayesian Approach for Combining Information from Multiple Surveys in Small-area Estimation Using Public-use Data
|
|
Author(s):
|
Dawei Xie*+ and Trivellore E. Raghunathan
|
|
Companies:
|
University of Michigan and University of Michigan
|
|
Address:
|
2151 Hubbard, #11, Ann Arbor, MI, 48105,
|
|
Keywords:
|
posterior predictive distribution ; constrained estimates
|
|
Abstract:
|
Cancer surveillance research requires accurate county -evel prevalence rates of screening and risk factors. Raghunathan et al. (2003) developed a hierarchical Bayesian approach for obtaining such estimates by combining information from in-house data from the Behavioral Risk Factor Surveillance System (BRFSS) and the National Health Interview Survey (NHIS). Due to confidentiality concerns, however, it is not clear whether these model-based estimates obtained using in-house data can be released to researchers for use in their research. We develop estimates at the MSA level using publicly available data and then use the estimated model, the county level covariates and the direct estimates from BRFSS to obtain county level estimates. A Markov chain Monte Carlo (MCMC) method is used to generate the joint posterior predictive distribution of the county level unknown quantities. Yearly county level estimates for 49 states, District of Columbia and the whole state of Alaska in 1997-2000 are developed and compared to the HB model estimates from the in-house NHIS and BRFSS data.
|
