Annual Meeting 2005

The purpose of this short course is to introduce participants to new methods in adaptive sampling, look at where these designs can be used, how they work, and enable course participants to use these methods. The idea of adaptive sampling is to do a more effective survey by focusing the sampling effort as you go, depending on what you find, and to take advantage of the configuration of the population to find your way into particularly informative regions. For example, in surveys of an unevenly distributed species, one can look more intensively near sample units with high abundance or interesting values, while at the same time striving to fill in blank spots. Similarly, health studies of hard-to-survey human populations can be enhanced by tracing social connections to find more people of interest to the study.

The new adaptive designs and inference methods described in this short course have advantages over previous methods, such as flexibility in implementation, control of sample sizes, and better distribution of effort. Basic familiarity with statistics is assumed but no prior expertise in sampling techniques is required.

1. Introduction adaptive designs spatial and network settings; environment, ecology, and health.
2. Active set adaptive designs examples: adaptive cluster sampling, snowball sampling, random walks uniform and targeted walks, adaptive web sampling.
3. Design based estimation simple estimates Rao-Blackwell improvements Markov chain resampling.
4. Model based estimation Bayes approach data augmentation and other MCMC methods.