Analysis and Interpretation of Freshwater Fisheries Data

3: Sampling and Experimental Design

Michael J. Hansen, T. Douglas Beard Jr., and Daniel B. Hayes

doi: https://doi.org/10.47886/9781888569773.ch3

Even though many different methods are used to sample fish populations, their habitats, and anglers, sampling plans often share common traits. The appropriate gear and protocol to collect data are important, but the sampling design and the characteristics of the population determine the statistical properties of the estimates obtained. Similarly, when experiments are conducted to determine the response of fishes or their habitats to treatment, the overall experimental design and underlying variability determine the power of the experiment and can limit the questions or hypotheses that can be addressed. The goal of this chapter is to describe some of the most common sampling and experimental designs used in fisheries science. Our principal intent is not to teach the theory underlying these topics but rather to illustrate common data analysis approaches based on that statistical theory.

Fisheries scientists take samples from populations because data or information from all individuals in the population typically cannot be obtained. Fundamental to the idea of sampling is that a population of sampling units exists from which samples are taken. Ideally, all sampling units in the population can be sampled, but in many field sampling programs the sample frame, or the set of sample units that are actually available to be sampled, may be only a subset of the entire target population. In general, whenever the sample units in the sample frame differ from the units in the target population, the design may provide results that reflect the sample frame but not the target population (termed bias; see section 3.1.2). The degree of bias due to this mismatch is generally case specific and is virtually impossible to determine. Throughout this chapter, we assume that the sample units in the sampling frame match the units in the target population and that all units are sampled with equal efficiency.