|
Activity Number:
|
534
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #307524 |
|
Title:
|
Confidence Interval Coverage in Complex Model-Based Estimation with Deep Stratification
|
|
Author(s):
|
Archana Joshee*+ and Wendy Rotz
|
|
Companies:
|
Ernst & Young LLP and Ernst & Young LLP
|
|
Address:
|
1225 Connecticut Ave., NW, Washington, DC, 20036,
|
|
Keywords:
|
deep stratification ; model based estimation ; degrees of freedom
|
|
Abstract:
|
This paper studies confidence interval coverage in model based estimation under complex designs with two layers of stratification: a major stratum designation and deep stratification within each major stratum. The paper explores overall degrees of freedom (df) and confidence interval coverage in different settings. It is known in design-based stratified estimation, under specific assumptions, the Satterthwaite approximation estimates the df to lie between the df of the smallest stratum and the sum of the strata df. This approximation is not robust in presence of positive kurtosis. Yet little research has been done to explore df and the robustness of confidence interval coverage in complex model based estimation. This paper explores the df and robustness of confidence interval coverage in complex sample designs with deep stratification within major strata.
|
