Robustness Properties of Confidence Intervals on the Parameters of the Lognormal Distribution
by
Arvind K. Shah
University of South Alabama
Coauthors: John J. Lefante, Jr. (Tulane University)

In most applications involving positively skewed data, the lognormal distribution is assumed for the underlying population. Standard inference procedures exist and are widely used for estimating parameters of the lognormal distribution. Some common areas of application include modeling in environmental and occupational exposure settings and estimation of wildlife and marine populations. Often the assumption of the lognormal distribution is made in practice when in fact the underlying distribution may differ from a lognormal. Little has been reported on the effects departures from the assumed lognormal distribution have on inferences made when using lognormal dependent estimation methods. This paper investigates the robustness properties of the various exact and approximate lognormal confidence intervals, for estimating the mean. The gamma distribution is used to represent departures from the lognormal distribution. The effects this departure has on the width and coverage of various confidence intervals are investigated through simulations, by varying mean, variance, and sample size.

Date received: October 19, 2000

Copyright © 2000 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cafr-69.