Threshold selection in univariate extreme value analysis

Abstract Threshold selection plays a key role in various aspects of statistical inference rare events. In this work, two new threshold methods are introduced. The first approach measures the fit exponential approximation above and achieves good performance small samples. second method smoothly estimates asymptotic mean squared error Hill estimator performs consistently well over wide range processes. Both analyzed theoretically, compared to existing procedures an extensive simulation study applied dataset financial losses, where underlying extreme value index is assumed vary time.