Statistics & probability letters vol:60 issue:3 pages:265-278
The estimation of a dependence index introduced in bivariate extreme value methodology by Ledford and Tawn (Biometrika 83(1) (1996) 169) is discussed. It is argued that estimators with bad bias properties are to be avoided. In this spirit we also suggest a new estimator. (C) 2002 Elsevier Science B.V. All rights reserved.