Content area
Abstract
Determination of exact values of certain population parameters is often infeasible. As a result, we use statistics derived from random samples to produce an estimate of the actual population parameter. These estimates are then used to construct a confidence interval (C.I.) within which we expect the actual population parameter to fall. We find that as the sample size is varied, the actual coverage of the C.I.'s will sometimes be significantly different from the theoretical value especially in the case of smaller sample sizes, n<100. In this thesis, we investigate various methods of constructing confidence intervals for certain functions of the mean and variance of a population. Since the observed coverage is largely dependent on the population distribution, sample size used and the method used to create the confidence interval, we are interested in determining the method that produces the best confidence interval using a given sample of size n from a specific population.
