Estimation and testing of quantiles of the extreme-value distribution (Q1085922)

This paper deals with the ABLUE (asymptotically best linear unbiased estimators) and a test of the hypothesis quantile Q(\(\xi)\) \((0<\xi <1)\) of the location-scale parameters of the extreme-value distribution for maxima (largest value) based on k (\(\leq n)\) optimally chosen order statistics in a large sample of size n. Tables of optimum spacings, coefficients of the ABLUE, and the ARE (asymptotic relative efficiency) of the estimates compared to the corresponding full sample maximum likelihood estimators are provided for \(k=2(1)10\) and \(\xi =0.01\), 0.05(0.05)0.95, 0.99. It is shown that the same spacings may be used for both testing and confidence intervals.