Large deviations of tail estimators based on the Pareto approximation

Abstract
We consider the relative error of a tail functionwhen this is approximated byy–αusing an estimator of Hill's forα.The results combine recent work of Davis and Resnick on tail estimation with Anderson's work on large deviations in extreme-value theory. Treating separately the domains of attraction of Φαand Λ, we obtain general conditions for the relative error to tend to 0 asu→∞,y →∞ simultaneously. The results serve as warning against the automatic extrapolation of estimates based on extreme-value approximations.

This publication has 14 references indexed in Scilit: