Abstract: Modeling conditional dependency structure of financial assets through time-varying covariance is of great significance to capital risk management, option pricing and optimal portfolio. With the advent of high-frequency data, realized multivariate volatility model is widely applied in the financial field in recent years. Among the observation-driven models for time-varying parameters, the clear advantage of the generalized autoregressive score (GAS) model is that it exploits the complete density structure and likelihood information rather than theconditional moments information.

Key words: GAS-SKST-F model, multivariate skewed student t density, multivariate volatility