The Hill estimators under power normalization

By using the link between the affine and the power norming, eight Hill estimators under power normalization for the tail index (the non-zero extreme value index) are suggested. Moreover, more compact and adaptive four Hill estimators under power normalization are derived based on the generalized Pareto distributions under power normalization. Two classes of harmonic t-Hill estimators under power normalization are also suggested. A comprehensive simulation study using the R-package shows that all the suggested estimators under power normalization work well, but in all cases the Hill estimators under power normalization based on Pareto distributions under power normalization are better. The two models under linear and power normalization for extreme value analysis are applied with comparison on a real data set of two pollutants, Sulphur Dioxide and Particulate Matter.