Simple Random Sampling (SRS) is the most basic sampling technique. More complex sampling methods are derived from it. Since SRS requires enumerating the population, it cannot be applied to bulk materials. While cluster sampling is a substitute, classical statistical theory often requires knowing both the population size and the sample size. Groups of units are often assumed to be of equal size or volume. These requirements and assumptions do not hold for bulk sampling. Random sampling in this context is a fallacy.
For bulk material, Pierre Gy generalized the concept of random sampling by introducing the idea of "correct sampling." Three issues are involved. The sample has to be defined and extracted, and its integrity must be preserved. These concerns do not arise in SRS. In bulk sampling, however, their presence introduces bias, increases variation, and produces unpredictable results. If samples are not correct, then the statistical modeling and computations on the resulting data are not valid.
Several sampling systems and procedures will be examined for correctness. Gy's theory will provide the basic principles for identifying and reducing the sampling "errors."
|