Some comments on the estimation of a dependence index in bivariate extreme value statistics