Activity Number:
|
302
|
Type:
|
Topic Contributed
|
Date/Time:
|
Wednesday, August 14, 2002 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Social Statistics Section*
|
Abstract - #300968 |
Title:
|
Two-stage Sample Design with Small Clusters
|
Author(s):
|
Robert Clark*+ and David Steel
|
Affiliation(s):
|
University of Wollongong and University of Wollongong
|
Address:
|
Northfields Avenue, Wollongong NS@, International, 2522, Australia
|
Keywords:
|
sample design ; household surveys ; telephone surveys
|
Abstract:
|
In two-stage surveys, the sample sizes of clusters and units within clusters can be chosen to minimise the variance of an estimator, for fixed cost. If there are only a small number of units in each cluster, as in household surveys, then the optimization should be over integers. This paper considers sample designs where the number of units to be selected from each cluster is a function of the cluster size. An integer programming method is developed, which gives significantly lower variances than traditional methods. A non-integer within-cluster sample size can be implemented by using a mixture of several integers; this can further reduce the variance.
|