Further Development of New Methods for Estimating Tail Probabilities and Extreme Value Distributions

摘要

This research has focused on the problem of obtaining confidence intervals for extreme quantiles based on a random sample form a distribution of unknown form. Three confidence interval procedures were studied, both analytically and by means of an extensive Monte Carlo experiment. The experiment involved three sampled sizes (100, 200, 400) and twenty underlying distributions (five Weibulls, five mixed Weibulls, five lognormals, five mixed lognormals). The Monte Carlo results show that all three procedures studied work quite well and point the way to further improvement. (Author)