Session Slot: 2:00- 3:50 Sunday
Estimated Audience Size: 50
AudioVisual Request: none dws
Session Title: Network and Adaptive Sampling-Theory and Application
Conventional sample designs uniquely link only one sampling unit to each population element. With the development in the 1970's of network sampling and the concept of a counting rule, survey researchers at the design stage had available a choice of counting rules that allowed more than one sampling unit to report a population element. Network sampling has been shown to be useful for reducing: (1) sampling errors for estimating population quantities for rare and elusive populations, (2) measurement biases for coverage deficiencies and reporting errors, and (3) dependence in dual system estimation. Network sampling has been applied to many difficult problems such as estimating the prevalence of illicit drug use, diabetes, and cancer, and the undercount in the US Census. Another type of sampling useful for surveying rare populations is adaptive cluster sampling. In adaptive cluster sampling, an initial probability selection of units is made and then after the sampled units with the attribute of interest are identified additional units in a neighborhood of these units are sampled. The neighborhood often consists of sampling units that are geographically close to each other. This type of sampling can be quite efficient when there is geographical clustering of the population of interest. For example, in ecological applications where there is interest in estimating the population size of rare species of plants or animals which tend to grow or live in close proximty to each other.
There are interesting connections in the application and theory of
network and adaptive cluster sampling which can be explored. The purpose of
this session is bring together practitioners and methodologists working on
problems using both types of sampling so that they can present
applications and methodological developments.
Theme Session: Yes/No
Applied Session: Yes
Session Organizer: Graubard, Barry National Cancer Institute
Session Timing: 110 minutes total (Sorry about format):
110 minutes total...please allocate Opening Remarks by Chair - 0 minutes First Speaker - 25 minutes Second Speaker - 25 minutes Third Speaker - 25 minutes Fourth Speaker - 25 minutes Floor Discussion - 10 minutes
Session Chair: Graubard, Barry National Cancer Institute
1. The History of Network Sampling
Sirken, Monroe, National Center for Health Statistics
Abstract: This paper reviews the history of network sampling, and the applications of network sampling in circumstances where multiple selection units are naturally linked to the same population elements, and other circumstances where multiple linkages are deliberately fostered to improve survey design efficiency. Also, the paper describes the network sampling paradigm, and the survey counting rule and other important design features of network surveys.
2. Population Based Network Sample Surveys of Establishments
Shimizu, Iris, National Center for Health Statistics
Address: 6525 Belcrest Road, Hyattsville, MD 20782
Sirken, Monroe, National Center for Health Statistics, Westat
Judkins, David R., National Center for Health Statistics, Westat
Abstract: In a population based network sample survey of medical providers (vis. hospitals, physicians, dentists, clinics, etc.), the sample providers are selected from among those reported as being seen by respondents in a household sample survey. The network sample design is suggested whenever good medical provider sampling frames are unavailable. This paper discusses other conditions favoring the network design as well as other factors that may contraindicate its use.
3. Adaptive Sampling in Graphs
Thompson, Steven K., Pennsylvania State University
Abstract: Adaptive sampling designs are those in which the procedure for selecting the units to include in the sample may depend on values of variables of interest observed during the survey. For example, neighboring units may be added to the sample whenever high values are observed. In spatial sampling the neighborhood is defined by geographic proximity. In studies of human populations the neighborhood may also be defined by social relationships.
In studies of hidden and hard-to-reach human populations such as injection drug users and others at risk for HIV transmission, adaptive link-tracing designs in which initial respondents lead investigators through social links to other individuals often provide the only practical way to obtain a sample large enough for the study. Data summaries or inference from such samples can be misleading, however, if the sample-selection procedure is not taken into account. The situation is conceptualized as sampling in a graph, with the nodes of the graph representing people and the arcs or arrows represeting social relationships. The problem is that data are observed for only a sample of the nodes and arcs, from which we wish to infer characteristics of the whole graph or population.
Examples of link-tracing designs include network sampling, snowball sampling, chain-referral methods, ``random walk'' designs, and adaptive cluster sampling. Design-based and model-based methods of inference with such designs will be discussed in this talk.
4. Using a Restricted Adaptive Sampling to Estimate Fish Larval Abundance
Lo, Nancy C.H., Southwest Fisheries Science Center
Address: P.O. Box 271, La Jolla CA 92038
Griffith, David, Southwest Fisheries Science Center
Hunter, John R., Southwest Fisheries Science Center
Abstract: Adaptive sampling is a sampling design in which the procedure for selecting sample sites and allocating sampling effort depends on data collected during the survey. In March 9 - 27, 1995, a restricted stratified adaptive sampling was used to survey Pacific hake larvae in California water because the spatial distribution of Pacific hake larvae is highly patchy. The survey was conducted between Los Angeles and San Francisco covering an area of 202,115 km2 (59,540 nm2). Because of limited survey time, we used a restricted adaptive sampling design imposing a maximum number of station in each stratum. A stratified two-stage cluster Horvitz-Thompson (HT) and a simple stratified (SS) sample mean were used to estimate mean catch per tow. For our survey design, although HT is biased , the variance of the HT estimate was high. The variance of SS was lower than the variance of sample mean from both simple random sampling and proportional stratified sampling. Simulation confirmed our conclusions. It also estimated bias of SS, and showed that the sample variances of HT and SS were biased downward. In some cases, SRS performed well. Nonetheless, our adaptive sampling was relatively easy to implement, and it provided biological information within patches, which may not have been possible if conventional sampling designs were used.
List of speakers who are nonmembers: None