For five popular parametric copulas, classical maximum-likelihood is compared to a total of nine different minimum-distance estimators. The purpose of this paper is to present a comprehensive simulation study on the finite sample properties of minimum-distance and maximum-likelihood estimators for bivariate and multivariate parametric copulas. In the empirical risk management application, the practical usefullness of this strategy isĮxemplified for a set of bivariate portfolios. Strategy improves the power of GoF-testing when used to identify the main component of a mixture copula. The results show that the exclusion of outliers can have a beneficial effect on the power of the GoF-tests. Multivariate outliers are applied to the contaminated data. In order to robustify the GoF-tests, several methods for the detection of The Monte Carlo simulations show that independent of the underlying true copula, the GoF-test or chosen test statistic, even minor contaminations of the data can lead to a significant decrease in the GoF-tests’ power. To assess the tests’ robustness, we consider perturbations and outliers both in the dependence structure and the observations from the joint distribution. This paper is the first to analyze the robustness of goodness-of-fit for bivariateĮlliptical and archimedean copulas.
0 Comments
Leave a Reply. |