Statistical sampling approaches

This article describes basic sampling principles and the application of statistical sampling techniques to specific problems encountered in the Food and Drug Administration (FDA). Concepts are emphasized, and theory is minimized. The basic principles of sampling from a normal and binomial population, including confidence interval calculation and sample size determination, are briefly reviewed. Stratified, random, systematic, and judgment sampling are explained. Operating characteristic curves for attribute (and perhaps variable) sampling for acceptance of lots are derived and applied to specific FDA problems. The advantages and disadvantages of single and multiple sampling plans and plans which address multiple classes of criteria such as major and minor defects are discussed. Sampling schedules such as MIL-STD-105D and Canada's Government Specifications Board CGSB-105-GP-1 are reviewed to familiarize readers with the principles involved in these plans and to give them an idea of how they could be applied to FDA problems.