Gustav Alfelt: Modelling covariance matrix time series

In this talk, I will give an introduction to modelling covariance matrices in a financial time series setting. The Wishart process arises as a natural candidate for the stochastic component of such models. As such the Conditional Autoregressive Wishart model will be presented, in the context of serially correlated data. Additionally I intend to discuss how the fit of Wishart process models can be evaluated with the aid of matrix decompositions.