Finding the tail of a distribution: analysis of a method based on the coefficient of variation

Estimation at extreme values focuses on the tail of the data distribution, where observations are scarce. One of the key issues in inference on extreme values is knowing where this tail begins. This is a widely studied topic in the literature, with various methods available for this purpose. This study aims to investigate one of these methods presented in the literature. The performance of the method will be evaluated based on a computational simulation study. We identify strengths and/or constraints of the method which may be useful for users. The method will be applied to real data, with the goal of estimating parameters of rare events, like the risky measures, Value-at-Risk and Expected Shortfall.