On the limiting behavior of the Pickands estimator for bivariate extreme- value distributions (Q1186650)

Possible nondegenerate limit laws for the extreme values of an i.i.d. sequence of bivariate random vectors are related to a bivariate extreme value distribution with survival function \[ H(x,y)=\exp\{ - (x+y)A(x/(x+y))\},\quad x,y\geq 0. \] For the dependence function \(A(u)\) defined on \(0\leq u\leq 1\), \textit{J. Pickands} [Bull. Int. Stat. Inst. 49, No. 2, 859-878 (1981; Zbl 0518.62045)] proposed an estimator. His estimator, based on a random sample \((X_ i,Y_ i),\;i=1\) to \(n\), is given by \[ A_ n(u)= n/[ \sum_{i=1}^ n\min \{X_ i u^{-1},Y_ i (1-u)^{-1}\}] . \] Let \(\delta_ n(u)=\sqrt n \left[ \bigl( A_ n (u)\bigr) ^{-1} - \bigl( A(u)\bigr) ^{-1}\right] \). Using general results for sums of random variables taking values in Banach spaces, the author establishes the functional central limit theorem and the law of the iterated logarithm for \(\delta_ n(u)\). These limit results are used to suggest a modified estimator of \(A(u)\) and a test for independence of \(X\) and \(Y\).