Dynamic Network Sampling

点击翻译按钮获取中文摘要

The proposed research develops new sampling strategies, termed dynamic network sampling, that are needed for important situations that are not well served by conventional sampling and intervention strategies. Examples of such situations include interventions in hard-to-reach human populations to alleviate the spread of health and social problems, controlling the spread of epidemics in animal, plant, and human populations, adaptive management of renewable resources, and moving sensor designs for detection of airborne pathogen releases, and pest management. The proposed sampling designs extend earlier designs of the investigator, including adaptive sampling, adaptive cluster sampling, targeted random walk network designs and adaptive web sampling, to the cases in which the population and the sampling procedure are fully dynamic. A dynamic network sample spreads itself through the changing population of interest by adaptively following links in networks and by using distance relationships in spatial environments. The selection probabilities for link tracing depend adaptively on values of sample and one-step-away nodes, link characteristics, current sample size, and target sample totals, and other variables. The proposed sampling strategies can be used for for introducing interventions and experiments into hard-to-reach populations and for making inferences about population characteristics. Preliminary studies indicate the proposed sampling strategies are likely to be substantially more effective than conventional strategies for each of those uses.