|
Activity Number:
|
316
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #305213 |
|
Title:
|
Bootstrapping for Variance Estimation in Multilevel Models Fitted to Survey Data
|
|
Author(s):
|
Milorad Kovacevic*+ and Rong Huang and Yong You
|
|
Companies:
|
Statistics Canada and Statistics Canada and Statistics Canada
|
|
Address:
|
120 Parkdale Ave., Ottawa, ON, K1A0T6, Canada
|
|
Keywords:
|
informative design ; multi-stage design ; survey weights ; variance components
|
|
Abstract:
|
For estimation of parameters of a multilevel model fitted to hierarchical survey data, typically the survey weights and their bootstrap replicates are available only for the level-one units, thus making impossible implementation of the estimation methods that require weights at the higher levels. Some approximations are needed. In a simulation study designed according to the Canadian Workplace and Employee Survey, we study the impact of this limitation on the properties of mixed-model parameters estimates and their standard errors estimated by the survey bootstrap method. Several existing methods are compared and alternatives suggested.
|