Seminar by Dr. Boris CHOY from Discipline of Business Analytics, University of Sydney

In this talk, a class of non-elliptical multivariate probability distributions is proposed. These distributions are derived from the mean-variance mixtures. They are heavier-tailed than the multivariate normal distribution and offer very flexible tails. Pairwise linear correlation coefficients can be derived analytically. The mean-variance mixture of the probability density function allows effective model implementation using Markov chain Monte Carlo or expectation-maximization algorithms, identification of potential outliers and protection of statistical inference from the distorting effect of the outliers. In the empirical study, loss data in the form of run-off triangles are modelled using an ANOVA-type model with this class of non-elliptical error distributions. It is shown that this class of non-elliptical distributions outperforms its elliptical counterparts in stochastic loss reserving.

Dr. Boris Choy is an Associate Professor in Discipline of Business Analytics of the University of Sydney Business School and is a qualified statistician. He received his PhD in Statistics from Imperial College London, MPhil in Statistics from the Chinese University of Hong Kong and BSc. (First Class Honours) in Mathematics from the University of Leeds. His main research interests focus on robust statistical analysis of financial time series data and insurance data using heavy-tailed distributions via Bayesian computational methods and scale mixtures density representation.

Dr. Choy has received a number of teaching awards including the Vice-Chancellor Award for Outstanding Teaching from the University of Sydney.