Studies of drug users and other elusive populations are impeded by inherent sampling and estimation difficulties. On the one hand, probability-designed sample surveys tend to be expensive, produce a low yield of individuals with rare characteristics of interest, and are inefficient when populations are unevenly distributed. On the other hand, ethnographic and other in-depth studies of drug users and other elusive populations characteristically obtain a large enough sample only through link-tracing referral methods and statistically sound methods have not been available for making inferences from such samples to the larger population of interest. In this project, research is proposed to develop improved sampling design and estimation methods for studies of drug users and other elusive populations.
The aims of the proposed research include (1) development of statistically sound estimation methods for use with the link-tracing procedures prevalent in studies of drug users and other elusive populations, (2) development of improved sampling designs for studies of such populations, (3) development of sampling methods that combine traditional sample survey methods with adaptive or link-tracing procedures, (4) development of sampling methods to increase the yield in the sample of individuals with a rare characteristic, and (5) investigation and development of methods for dealing with non-sampling errors in surveys of drug users and other elusive populations. The proposed methods include the use of adaptive and graph sampling strategies and utilize a variety of design-based and model-based approaches to sampling and estimation.
| Thompson, Steven K; Collins, Linda M (2002) Adaptive sampling in research on risk-related behaviors. Drug Alcohol Depend 68 Suppl 1:S57-67 |
