|
Activity Number:
|
215
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 10, 2004 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #301824 |
|
Title:
|
Small-area Estimates of Diabetes and Smoking Prevalence for North Carolina Counties: 1996-2002 Behavioral Risk Factor Surveillance System-based Estimates
|
|
Author(s):
|
Akhil Vaish*+ and Neeraja Sathe and Ralph Folsom
|
|
Companies:
|
RTI International and RTI International and RTI International
|
|
Address:
|
3040 Cornwallis Rd., RTP, NC, 27709,
|
|
Keywords:
|
survey-weighted hierarchical Bayes ; pseudo-score functions ; MCMC NC county-level SAEs for diabetes and smoking ;
|
|
Abstract:
|
We present North Carolina (NC) county-level Small Area Estimates (SAEs) for diabetes and smoking. The pooled 1996-2002 NC-Behavioral Risk Factor Surveillance System (BRFSS) data for diabetes and smoking is used to produce SAEs for all the 100 counties in NC. Estimates of change in diabetes and smoking prevalence rates from 1996-1999 to 2000-2002 are also obtained. The Survey Weighted Hierarchical Bayes (SWHB) methodology of Folsom, Shah, and Vaish (1999) is used to fit mixed logistic regression models. Multivariate age group specific random effects defined at two levels of hierarchy along with personal level covariates (race and gender) as well as county-level predictors obtained from various national data sources are used in the SWHB modeling. The SWHB methodology has several benefits over commercially available software such as MLwiN and BUGS. The SWHB methodology allows the use of personal-level predictors in the modeling and as a result SWHB-SAEs are internally consistent and more efficient than the Fay-Herriot (1979) aggregate level solution. The SWHB-SAEs are also design consistent and self-benchmarked to the robust survey-weighted estimates for large sample areas.
|
