|
Abstract:
|
In occupational health studies, it frequently happens that the measurements are lognormally distributed , but because government standards are based on arithmetic means, appropriate means and confidence intervals must be estimated. Past work on the confidence interval problem has mostly been devoted to simple random samples from lognormal populations. The methods used include the bootstrap and generalized confidence intervals. When confidence intervals are required for the arithmetic mean of a factor in a mixed or random effects linear model, the methods from simple random sampling may be used by grouping the data to obtain random samples. However, possible complications arise when confidence intervals must be provided for averages of arithmetic means, for which results are given. This work compares results obtained by the above methods.
|
