Use of the Lognormal Distribution for Survival Data: Inference and Robustness

Two data sets are presented and various distributions, including the lognormal, are fitted to the data. A method is given to calculate exact confidence intervals for the quantiles of the lognormal distribution. The coverage probability of the confidence intervals is examined when the lognormal distribution is the correct model, and for various departures from lognormality. In addition, the connection between the coverage probability and the pvalue from a goodness-of-fit test is explored.