A0411
Title: Tractable unified skew-t distribution and copula: Specification, inference and application
Authors: Lin Deng - University of Melbourne (Australia) [presenting]
Michael Stanley Smith - Melbourne Business School (Australia)
Ole Maneesoonthorn - Monash University (Australia)
Abstract: The unified skew-t (UST) distribution is a promising generalization of the popular multivariate skew-t distribution proposed by a prior study. It allows for greater flexibility in the asymmetry of the distribution and range of asymmetric dependence in its implied copula. However, estimation is difficult because it suffers from parameter identification and computational hurdles, inhibiting its use for modeling data. A tractable variant of the unified skew-t (TrUST) distribution and its implied copula are proposed. Some of the key properties of both are outlined, as is the Bayesian methodology for their estimation that uses an extended likelihood of a numerically stable generative representation of the TrUST distribution. The flexibility of both the TrUST distribution and its implied copula, along with the efficacy of the Bayesian inference, are first illustrated using simulated data. An application of the TrUST distribution to Australian electricity prices demonstrates the advantages of its increased flexibility. An application of the TrUST copula to high-frequency intraday equity returns shows it captures heterogeneity in asymmetric dependence over equity pairs more accurately than the popular skew-t copula.
