|
Activity Number:
|
389
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #309314 |
|
Title:
|
Using Bootstrap Variance Calculations for a Survey with a Simple Design: The Case of the 2005 National Survey of the Work and Health of Nurses
|
|
Author(s):
|
Gildas Kleim*+ and Yves Bélanger
|
|
Companies:
|
Statistics Canada and Statistics Canada
|
|
Address:
|
18-A R.H. Coats building, Ottawa, ON, K1A 0T6, Canada
|
|
Keywords:
|
Variance ; Bootstrap ; Simple Design
|
|
Abstract:
|
Rao and Wu (1988) proposed a bootstrap method aimed for use in complex sample designs involving stratification, multiple stages and unequal probability sampling. They also presented a bootstrap procedure for stratified simple random sampling without replacement (WOR). Even though formulas for variance estimates for such a simple design are well known, it may happen nevertheless that clients request bootstrap weights to be produced. The usual bootstrap procedure used by Statistics Canada, and referred to as the Rao-Wu bootstrap, assumes sampling with replacement (WR) and thus leads to an overestimation of the variance for WOR sampling, especially when the stratum sampling fractions are high. In this paper, we describe and evaluate an implementation of the bootstrap procedure for a simple design using the 2005 National Survey of the Work and Health of Nurses.
|
