R code used in the master thesis "Estimation of the Dependence Parameter in Bivariate Archimedean Copula Models Under Misspecification"

Copulas have become increasingly popular in multivariate statistics and financial applications. This paper studies the effect of misspecification among the three Archimedean copula families Frank, Gumbel, and Clayton on the dependence parameter estimation for two dimensions. In addition to the maximum likelihood estimator and the inverted Kendall's tau estimator, a p-value weighted average of the two is proposed and studied. To assess the performance of the proposed estimator, a comprehensive simulation study was conducted. As opposed to ML, the suggested estimator is shown to yield unbiased results even under copula misspecification for certain combinations of true copula, misspecified copula and dependence level. In the given application of estimating the Value-at-Risk of two bivariate portfolios using the three Archimedean copulas in combination with each of the three estimators, the proposed estimator also outperformed the ML estimator on the whole.