Proposal of a Method for Analyzing Operational Risk Loss Data

Abstract

Abstract. We propose a novel method for analyzing operational risk loss data. In this method, the marginal distribution of multidimensional operational risk loss data are formulated by a discrete distribution, the dependency is formulated by a D-vine t-copula, and parameters are estimated simultaneously with Markov chain Monte Carlo(MCMC) via the Bayesian method. Numerical experiments confirmed the practicality of the proposed method.