On the estimation of the extreme-value index and large quantile estimation (Q913399)

The authors consider an estimation problem on the main parameter of an extreme-value distribution, and discuss the asymptotic properties of Pickands' estimator based on the ascending order statistics of a sample. They give a short proof of the weak consistency of the estimator and natural conditions under which the estimator is strongly consistent and asymptotically normal. They obtain a result which enables one to construct a confidence interval for a high quantile and an endpoint of the distribution and also present some simulation experiments and an application of the results to the high tide water levels at the Dutch island Terschelling.