Number of test samples needed to obtain a desired Bayesian confidence interval for a proportion.

Abstract

One recurring problem in military operational test and evaluation is determination of the number of items to test. This thesis describes a Bayesian method to determine the sample size that is needed to estimate a proportion or probability with a (1-ą)100 confidence when a prior distribution is given to that proportion. It uses the two variants of the triangular distribution as priors and develops computer programs, graphs, and tables to assist in finding the required sample size. These results are compared with other approaches in determining the required sample sizes that are needed to obtain a desired confidence interval for a proportion or probability.