CONFIDENCE-INTERVALS FOR ARITHMETIC MEANS OF LOGNORMALLY DISTRIBUTED EXPOSURES

Distributions of concentrations of environmental contaminants are often approximately lognormally distributed. The arithmetic mean of such distributions is frequently the summary measure most relevant to health. Although methods for estimating arithmetic means of lognormal distributions from samples are becoming better known, methods for determining the precision of estimates by confidence intervals are not. In this paper, four approximate intervals are compared with the computationally cumbersome exact interval. Some of the approximate intervals are adequate for larger sample sizes (greater-than-or-equal-to 5), or small geometric standard deviations (GSD less-than-or-equal-to 1.5), but none of them are accurate for small sample sizes and large GSDs. A table presents exact 95% confidence intervals for small sample sizes with GSDs up to 4.0. Real examples illustrate the methods.