Errors Incurred When Using Iwao's Sequential Decision Rule in Insect Sampling

Abstract

Sequential decision rules based upon defining a relationship between the mean and variance of a population and using this relationship in a confidence interval about a sum of counts on sample units have been developed and used for sampling insect pests. The actual error rates that arise from the use of these rules are much larger than the nominal ones. This occurs for three reasons: The confidence interval is incorrectly used as a sequential decision rule; it is assumed that random variables are normally distributed; and variances are not constant. If these sequential decision rules are to be used, we suggest that actual error rates be determined through simulation.