An Introduction to Simple Latin Square Sampling and Adaptive Cluster Sampling
Dr. John Borkowski (Dept. of Mathematical Sciences, MSU)
4/24/24:
The final statistics seminar for the semester will be by our very own Dr. John Borkowski. Join us to hear from Dr. Borkowski who is retiring at the end of this semester after an outstanding career here at MSU!
Abstract: If spatial correlation is present, sampling designs which ensure the sample is well-distributed over the study region will, in general, improve estimation of population abundance relative to designs that do not. Designs based on simple latin square sampling (SLSS) will be presented that address this issue. When the goal is estimation of the population total or mean, these designs when compared to simple random, stratified, and systematic sampling will, in general, provide estimators with smaller variance when spatial correlation is present. In this talk I will provide an introduction to SLSS-based designs.
Next, with an adaptive sampling scheme the procedure for selecting units to include in a sample may depend on values of the variable of interest as it is observed during the implementation of the survey. That is, the sampling protocol is "adapted" the data. Adaptive cluster sampling (ACS) is a specific type of an adaptive sampling scheme. Unlike classical sampling plans, ACS operates under the rule that if an observed value of a sampling unit satisfies some condition of interest C, then additional units in its neighborhood are adaptively added to the sample. Then, if any of the additional units satisfy C, units in their corresponding neighborhoods are added to the sample as well. This adaptive process continues until no additional units satisfying C are encountered. The primary application of ACS is when sampling from populations whose members are found in spatially-aggregated clusters (e.g., herds of animals, nesting birds, ...), and the goal is to estimate a population total or mean. In this talk I will provide an introduction to ACS.