TABLE OF CONTENTS When a reliability or a life test is to be carried out, a sample of size N is taken from the population, and the required statistics and/or distribution parameters are determined. But, in practice, the first question that has to be answered before a sample is taken is "How large should the sample size, N, be?"
There are many ways of determining the size of the sample. For example, if a fixed sum of money has been allotted to carry out the test, then N may be as large as the allotted money allows, or it can be determined using a sample-size-cost model from which the minimum cost sample size can be calculated. However, the method often used in practice is to specify the size of the error that is desired in determining the statistic or parameter, or the Type I and Type II errors, and then to calculate the sample size from these quantities. Two kinds of sample size determination are discussed here: (1) Estimation, the object being to find N such that the true but unknown parameter is contained in a specified confidence interval with probability (1 - ?). (2) Testing hypothesis, the object being to find N such that the requirements for Type I ( ?) and Type II ( ?) errors are satisfied. The distributions covered here are the normal, the exponential and the Weibull.
