Statistical power estimation in Extreme Value Theory

Published: 29 January 2018| Version 1 | DOI: 10.17632/hh2byrbbmf.1
Federico Reghenzani,


These files contain the code (three executables in total) and the results carried out by that code. It provides the statistical power analysis and its sensitivity for Kolmogorov-Smirnov (KS), Anderson-Darling (AD) and the Modified Anderson-Darling (MAD) tests for Extreme Value (EV) distributions. The power is estimated via a Monte Carlo approach with 10^9 samples, under "case 0" conditions. The tests were performed varying the reference distribution, the sampling distribution, the sample size and the critical value. The sensitivity analysis was performed on the shape parameter of EV distribution.



Politecnico di Milano Dipartimento di Elettronica Informazione e Bioingegneria


Real-Time Systems, Statistical Hypothesis Testing, Extreme Value Theory