Gaißer, Sandra Caterina (2010) Statistics for Copula-based Measures of Multivariate Association - Theory and Applications to Financial Data. PhD thesis, Universität zu Köln.
Concepts of association or dependence play a central role when considering multiple random sources in statistical models as they describe the relationship between two or more random variables. In particular, the concept of copulas has proven to be useful in many fields of application and research. Copulas split the multivariate distribution function of a random vector into the univariate marginal distribution functions and the dependence structure represented by the copula. This dissertation addresses the modeling, the estimation and the statistical inference of multivariate versions of copula-based measures of association such as Spearman's rho. Special focus is put on the analysis of the statistical properties of related nonparametric estimators as well as the derivation of statistical hypothesis tests. The latter may be used to verify specific modeling assumptions such as, for example, equal pairwise rank correlation. Further, statistical tests are developed to identify significant changes of association over time. The theoretical results are illustrated with applications to financial data.
Actions (login required)