2023년 12월 8일 진행

개요:

This presentation introduces various methods for detecting multiple change points in inhomogeneous time series using asymmetric norms. Initially, the concept of an asymmetric norm is presented. Then a quantile segmentation method based on a moving window is demonstrated. The proposed approach enhances the moving window framework by employing quantile differences as test statistics and controlling family-wise error rates through the derivation of the null distribution. The method also allows for multiscale segmentation by merging candidates obtained from various bandwidths. Change point analysis of extremal distribution, which was suggested by Kojadinovic and Naveau (2017), is also presented, applied on the extremal precipitation of South Korea.