|
Activity Number:
|
386
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 15, 2002 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods*
|
| Abstract - #301776 |
|
Title:
|
Reweighting a National Database to Improve the Accuracy of State Estimates
|
|
Author(s):
|
Allen Schirm*+ and Alan Zaslavsky
|
|
Affiliation(s):
|
Mathematica Policy Research, Inc. and Harvard University
|
|
Address:
|
600 Maryland Avenue, S.W., Suite 550, Washington, District of Columbia, 20024-2512, USA
|
|
Keywords:
|
Small Area Estimation ; Simulation ; Weighting ; Current Population Survey
|
|
Abstract:
|
Policy analysis often relies on small area estimates for the many cells of many tables. Because sample sizes for small areas are typically inadequate to provide usable direct estimates, a model-based approach is needed. However, it is impractical to develop Empirical Bayes or other indirect estimators for every required table cell. Instead, we can reweight a national database by fitting a Poisson regression model to obtain an estimated prevalence in each small area of every household in the database. The model controls important aggregates for each area to match direct or smoothed (Empirical Bayes) estimates. Estimated prevalences are expressed as a matrix of weights, with each household having a weight for every small area. Once the weights are estimated, no further modeling is required. Any estimates sought for an area are obtained using all the households in the database and the weights for that area. Through extensive simulations, we have evaluated this method when the small areas are states. Our findings suggest that indirect estimates obtained using the reweighting method may be substantially more accurate than direct estimates.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2002 program |
