|
Activity Number:
|
537
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #306300 |
|
Title:
|
Using the t-Distribution in Small-Area Estimation: an Application to SAIPE State Poverty Models
|
|
Author(s):
|
Elizabeth Huang*+ and William R. Bell
|
|
Companies:
|
U.S. Census Bureau and U.S. Census Bureau
|
|
Address:
|
7005 Petunia Street, Springfield, VA, 22152,
|
|
Keywords:
|
Fay-Herriot model ; robust models ; current population survey ; American Community Survey
|
|
Abstract:
|
The Census Bureau's Small Area Income and Poverty Estimates (SAIPE) program produces state age-group poverty estimates from Bayesian treatment of a Fay-Herriot model applied to direct state estimates from the Current Population Survey (CPS). In recent years supplementary surveys for the American Community Survey (ACS) have also produced state poverty estimates. Huang and Bell (2004) compared posterior variances from the bivariate and univariate models for the CPS and ACS data. While we found some improvements with bivariate models under certain assumptions, we also found occasional large posterior variance increases. These corresponded to outliers or near outliers in the ACS equation. As a means of dealing with this "problem" we examine letting either the model errors or sampling errors in either the ACS or CPS equation follow a t-distribution.
|
