Title: Statistical properties of the modified multivariate skew-t distribution
Authors: Charles Au - University of Sydney (Australia) [presenting]
Boris Choy - University of Sydney (Australia)
Abstract: A class of distributions known as the multivariate skew-normal distribution has been a popular choice for capturing skewness and correlation in multivariate data. A heavy-tailed version of this distribution, known as the skew-t distribution, is often used for data with fat tails, such as asset returns. A more general extension to the skew-t distribution, the modified multivariate skew-t distribution (Mod-skew-t distribution), is proposed. It is more flexible in that the degrees of freedom parameter of each of the marginal distributions can be allowed to vary. This overcomes the limitation that these parameters must be the same, as is the case for the non-modified version of the skew-t distribution. Using the fact that the Mod-skew-t distribution has the scale mixtures of skew-normal (SMSN) representation, Bayesian inference will be used for parameter estimation. Its various statistical properties and application to statistical modelling will be explored.