CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Copula-based clustering of time series based on multivariate comonotonicity Authors:  Sebastian Fuchs - University of Salzburg (Austria)
Roberta Pappada - University of Trieste (Italy)
Fabrizio Durante - University of Salento (Italy) [presenting]
Abstract: In recent years, copula-based measures of association have been exploited to develop clustering methods that can take into account the dependence among different (one-dimensional) time series. In spatial statistics, such methods are particularly helpful in identifying hidden spatial patterns that define sub-regions characterized by a similar stochastic behaviour (i.e. regionalization). However, the majority of regionalization techniques focus on the spatial clustering of a single variable of interest, thus ignoring the role of compound events for extremes. Motivated by these problems, we propose a dissimilarity-based clustering procedure to group geographic sites characterized by multiple time series. In particular, the procedure tends to clustersites that exhibit a weak form of comonotonic behavior, which is more tailored for some applications. Different strategies to create such dissimilarity indices are hence illustrated and compared in a simulation study.