|
Activity Number:
|
133
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 7, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #306714 |
|
Title:
|
Small-Area Modeling for Survey Data with Smoothed Error Covariance Structure via Generalized Design Effects
|
|
Author(s):
|
Ralph Folsom*+ and Avinash C. Singh and Akhil Vaish
|
|
Companies:
|
RTI International and Statistics Canada and RTI International
|
|
Address:
|
3040 Cornwallis Road, Research Triangle Park, NC, 27709-2194,
|
|
Keywords:
|
estimating functions ; generalized deffs ; ignorable and nonignorable designs ; unstable estimated error covariance
|
|
Abstract:
|
We consider specifying the design-based error covariance structure in small-area modeling with survey data. While it is customary to treat the estimated covariance as known, it is often unstable. To alleviate this problem, one can either model the design-based covariance matrix or smooth the estimated covariance by specifying only its mean function. We prefer smoothing over modeling because of the strong assumptions needed to model the error covariance structure of SAEs. To smooth nondiagonal error covariance matrices, we make use of the g-deff (generalized design effect), defined earlier by Rao and Scott (1981) in the context of categorical data analysis. Simulation results for SAEs based on a linear mixed model show that the author's EFGL method (FCSM 2005) with the proposed smoothing provides improved coverage of confidence intervals.
|
