|
Activity Number:
|
170
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2003 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #301593 |
|
Title:
|
Model-based Optimal Selection of Sampling Units with Computational Ease
|
|
Author(s):
|
Chang-Tai Chao*+
|
|
Companies:
|
National Cheng-Kung University
|
|
Address:
|
Dept. of Statistics, Tainan, , , Taiwan
|
|
Keywords:
|
log-Gaussian spatial model ; model-based sampling ; optimal sampling strategy ; eigenvalue ; eigenvector ; Gaussian spatial model
|
|
Abstract:
|
An optimal model-based sampling strategy contains an unbiased estimator and a sampling design that selects sampling units to minimize the mean-square error. Under a given population model, one can find an optimal sampling strategy to minimize the mean squared error. In the previous results, the optimal sampling strategy usually involves intensive computation. The intensive computation usually soon becomes unaffordable as the population size increases. Hence, it is often not acceptable in practice. A practical procedure that utilizes the technique of multivariate analysis is proposed in this research. This procedure is able to find sampling units that often provide more efficient estimation than simple random sampling. The algorithm used in this procedure is fairly easy. The computation load does not increase too much when the population size increases. In addition, the proposed sampling strategy only needs the covariance structure but not the exact population distribution. The general selecting procedure will be described. The performance of this sampling scheme will be compared with simple random sampling under Gaussian and log-Gaussian spatial population models.
|
