A0581
Title: Analysis of variability in extremes with application in change point detection
Authors: Chen Yan - INRAE/Inria (France) [presenting]
Abstract: ANOVA is a widely used statistical method for comparing means across several groups. However, analyzing tail behavior rather than the mean can be more insightful in certain contexts. ANOVEX (ANalysis Of Variability in EXtremes) is introduced, a novel approach designed to compare extreme behaviors across J-1 groups. ANOVEX evaluates extreme quantiles within each group, utilizing methods like the Weissman estimator, and assesses the variances of extreme log-quantiles both within and between groups. Under the hypothesis of identical extreme behavior across groups, the variance ratio is shown to asymptotically follow a chi-square distribution with J-1 degrees of freedom. Additionally, ANOVEX's utility extends to change point detection and estimation in contexts divergent from traditional mean and variance shifts, particularly focusing on detecting and estimating changes in distributions' tails. This aspect broadens ANOVEX's applicability, offering a robust tool for analyzing extremes in various data-intensive fields.
