The Study On The Multivariate Volatility Time Series Model Based On Copula

In 1982, Engle presents conditional heterosoedastic model which aims at conditional two-order moments. Subsequently, conditional heterosoedastic model and stochastic volatility model has make a great progress which can capture some economic features such as volatility cluster. But the estimation of parameters for multivariate volatility model hampered its development.The problem has being resolved when copula theory appeared. As we all know, multivariate jiont distribution function clarify congruence between variables, which attracts our attentions when modlling a multivariate volatility time series. However, it is a complex multivariate function even it can't be established by it's marginal distribution which multivariate copula function can do.The thesis introduced multivariate copula function into modelling multivariate volatility time series model instead of conventional multivariate jiont distribution function such as multivariate normal distribution or multivariate student distribution. The degree and patten of dependence lying in models are systemic discussed in succession. At the same time, parameters of model are estimated and checked. The more interesting is we present a flexible mixed copula function which syncretize various abilities of copula function in expressing pattens of dependence,even different marginal distribution can be connected by a multivariate copula function to obtain a valid and flexible model.Such multivariate volatility copula time series model , which include more information but less parameters and offset the shortage of the conventional multivariate jiont distribution function assuption, not only can express degree and pattens of dependence flexibly and effectively but also strengthen the interpretative ability to datas.