|
Activity Number:
|
439
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #304748 |
|
Title:
|
Variance Estimates for Synthetic Estimation of Need for Mental Health Services
|
|
Author(s):
|
Hoang Nguyen*+ and Charles E. Holzer, III
|
|
Companies:
|
LifeStat LLC and The University of Texas Medical Branch at Galveston
|
|
Address:
|
2105 Silverthorn Lane, Kemah, TX, 77565,
|
|
Keywords:
|
BRR ; complex survey design ; synthetic estimation ; mental health ; NCS-R
|
|
Abstract:
|
Synthetic estimation is an approach that draws from a national survey of mental health (NCS-R) to create estimates of need for states and counties, through model based restandardization of the survey results. The variance estimates obtained directly from the original survey are changed by the models and the intrinsic weighting and projecting to specific geographic areas. To obtain the survey contribution to variance in the local estimates, we have used balanced repeated replication (BRR) of the entire modeling and application to the local areas to obtain estimates for those areas. Analysis of the small area variance of estimates indicates relatively small variance relative to the estimates themselves. We conclude that synthetic estimation is a valuable process for producing both state and county level estimates, compared to SAMHSA standard of using a single rate for all areas of the U.S.
|
