St. James Ballroom
A Sampling Strategy to Efficiently Estimate Proportions or Secondary Information About Objects of Interest in a Clustered, Rare Population and Accompanying R Package (303878)
Mary Christman, MCC Statistical Consulting, LLC*Kristen Erica Sauby, Lighthouse Statistical and Ecological Consulting, LLC
Keywords: adaptive cluster sampling, sampling efficiency, proportional data, interactions
The study of interactions (e.g., among plants and insects) often involves sampling to estimate proportions (e.g., the proportion of a particular host plant species infested by an insect) and may be quite costly because only a fraction of sampling locations with the first species will likely have the second species (e.g., the insect species). For studies involving populations of rare, clustered individuals, adaptive cluster sampling (ACS) can increase efficiency and reduce variance, compared to designs such as simple random sampling (SRS). However, the final sample size is unknown at the start of sampling and can be quite large if the sampled clusters are large. We propose a restricted ACS (RACS) design, where the number of sampled units in a cluster is limited to control the final sample size. We focus specifically on the design's ability to efficiently sample proportional information (e.g., the fraction of plants infested by insects). We present an ecological example to explore the design. We also present an R package, which allows for the evaluation of the suitability of the RACS design relative to ACS and SRS as well as visualization of the sampling process and results.