A0690
Title: Financial modeling under non-Gaussian distribution
Authors: Antonio Pacifico - University of Macerata (Italy) [presenting]
Abstract: The aim is to propose and develop a computational approach to improve the recent literature on multivariate GARCH (MGARCH) models. Generally, they consider relatively low-dimensional applications (N lower than or equal to 8, with N denoting the number of observations). Then, the conditional variance within each financial component series is modelled separately as in the standard univariate GARCH processes. However, the increasing volatility and uncertainty in financial markets during the recent global economic crisis and successive consolidation periods have confirmed the close linkage between financial data and the need to investigate multivariate stochastic volatility among multiple markets in a unified framework. The key steps in the proposed framework are as follows. First, a change-point methodology for (conditional) covariance structure of multivariate high dimensional MGARCH processes is proposed. Second, a Markov Chain through Metropolis-Hasting steps is constructed since it has the posterior distribution of the model parameters as a stationary distribution. Third, each MGARCH model is jointly evaluated through their posterior probabilities or Bayes factors for a given set of competing models. Finally, to obtain a sample of the joint posterior density of models and model parameters, MCMC implementations are addressed.
