Bias and Variability in Statistical Estimates
This chapter explores the fundamentals of variance and bias and identifies sources for each. For the pest management practitioner, sampling plans that are based on biased models will also be incorrect, but their full impact on management decisions is not clear. Bias in arthropod enumerations or presence/absence counts can occur in many ways, but it may be associated with one of three general sources: nonrandom sampling, sampling technique, and interpersonal error. Bias that arises from nonrandom sampling may be difficult to detect because the usual method of its estimation may not be available as it is difficult to know the true mean density or proportion of infested sample units in most agricultural situations. If the sampler finds that a selected sampling plan for a given arthropod yields a low level of precision, it may be possible to use another sampling plan that is known to produce biased results but provides an increased level of precision.