Designing an efficient adaptive cluster sample
by
Jennifer Brown
University of Canterbury

Adaptive cluster sampling can be an efficient design for surveying rare populations - if the survey is designed well. Two design factors are: i) value of the condition that triggers the adaptive selection of neighboring units, and ii) the definition of the neighborhood that is adaptively searched. The choice of these factors is critical for designing an efficient survey. However, to date, there are few practical guidelines on how to design such a survey. In designing an efficient survey one should aim for a small difference between the initial and final sample size, and a small differences between the within-network and population variances. However, the two aims can be at odds with each other because small differences between initial and final sample size usually means small within- network variance. <p> One way to help in designing an efficient survey is to think in terms of small network sizes since the network size is a function of both critical value and neighborhood definition. One should aim for networks that are small enough to ensure the final sample size is not excessively large compared with the initial sample size but large enough to ensure the within-network variance is a reasonable fraction of the population variance. In this study surveys that had networks that were two to four units in size were the most efficient.

Date received: May 26, 2002

Copyright © 2002 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cajj-11.