Cluster Sampling: A Pervasive, Yet Little Recognized Survey Design in Fisheries Research

Cluster sampling is a common survey design used pervasively in fisheries research to sample fish populations, but it is not widely recognized by researchers. Because fish collected via cluster sampling are not independent of each other, standard simple random sampling estimators and statistical tests that assume independence cannot be used to make inferences about fish populations. If the clustered nature of fisheries data is ignored, the main consequence is that the type I error rate of common statistical tests will be severely inflated and significant differences will often be found in group comparisons where none exist. The goal of this paper is to provide an introduction to the estimation of population attributes and analysis of fisheries data collected via cluster sampling. This article addresses the nature of clustered fisheries data, reviews the random cluster sampling estimators of population attributes, explores the implications of violating the assumption of independence in hypothesis testing, and reviews current statistical approaches that can be used to analyze appropriately clustered data.