LOCATION: 319 Snedecor
SPEAKER: Steve Thompson, Department of Statistics, Pennsylvania State
University, University Park, Pennsylvania and Statistical Sciences Group,
Los Alamos National Laboratory, Los Alamos, New Mexico
TITLE: Active Set Adaptive Sampling
ABSTRACT:
| Adaptive sampling designs are those in which the procedure for selecting the sample depends on values of variables of interest observed during the sampling. For example, in an aerial survey of a migratory waterfowl population having an unpredictably patchy spatial distribution, additional sample units may be selected in the vicinity of high observed abundance. In a survey of a hidden human population such as injection drug users or others at risk for HIV/AIDS, social relationship links are traced from respondents having high behavioral risk factors to add other members of the population to the sample. In this talk I'll describe adaptive sampling designs in which, at any point in the sampling, the next unit or set of units is with high probability selected from a distribution that depends on the values of variables of interest in an active set of units already selected. With some lower probability, the next selection is made from a distribution not dependent on those values. The active set may consist of the entire current sample, or only the most recently selected unit, or a wide range of other possibilities. Design-unbiased estimation with such designs is based on a combination of initial and conditional selection probabilities, and these preliminary estimators are improved using the Rao-Blackwell method. Markov chain resampling estimators are used for larger sample sizes. In comparison with other adaptive and link-tracing sampling methods, the present class of strategies has advantages in flexibility regarding adaptive criteria and breadth and depth of sample coverage, ease of implementation, control of sample sizes, and the availability of robust if computationally intense design-based estimators. COFFEE: 3:45 p.m., 104 Snedecor Hall |