In processes where solid material is fed into containers such as bags or drums at the end of the production line, it may not be logistically feasible or it may take too much time to wait until all containers are filled to select a random sample. The containers would have to be stored somewhere until all of them were filled. Then those selected for the sample would have to be retrieved. Or if the process were continuous and we shipped the product as soon as a truck could be loaded, then we would never have an entire lot from which to sample. The following method can be used for obtaining a random sample of discrete units when the lot items are accessible sequentially (Kennedy and Gentle, 1980).

Suppose we have a batch process in which bags are filled with product at the end of the line, and we would like to sample 1% of the bags. Then each bag should have probability 0.01 of being selected. Using a computer, we generate a series of random numbers ^[22] between 0 and 1, and we generate more numbers than we think there will be bags. Often, we do not know exactly how many bags there will be in a lot, but based on past experience we have a good idea of what the maximum number will be. We can even generate the random numbers on the spot, as would be necessary for a continuous process in which we keep filling and...