|
Activity Number:
|
133
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, August 12, 2002 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods*
|
| Abstract - #301392 |
|
Title:
|
On Incorporating the Selection Probabilities into Robust Model-based Inference
|
|
Author(s):
|
Roderick Little*+ and Hui Zheng
|
|
Affiliation(s):
|
University of Michigan and University of Michigan
|
|
Address:
|
1420 Washington Heights, Ann Arbor, Michigan, 48109-2029, USA
|
|
Keywords:
|
Survey Sampling ; Spline estimation
|
|
Abstract:
|
The Horvitz-Thompson (HT) estimator is a basic design-unbiased estimator of the finite population total for sample designs with unequal probabilities of inclusion. Viewed from a modeling perspective, the HT estimator performs well when the ratios of the outcome values y(i) and the selection probabilities pi(i) are approximately exchangeable. When this assumption is far from met, the HT estimator can be very inefficient. We consider model-based alternatives that posit a smoothly-varying relationship between y(i) (or a function of y(i)) and the inclusion probability pi(i) (or a function of pi(i)), and that model this relationship using P-splines. The methods are intended for situations with probability-proportional-to-size sampling and continuous survey outcomes. Design-based simulation studies show that spline-based predictive estimators clearly dominate the HT estimator when it performs poorly, with minor differences when the HT estimator performs well. Promising extensions to multistage sampling using P-splines with random effects are also presented.
|
