This thesis concerns estimation of stochastic volatility models in
finance, discrete time models as well as continuous time
Returns of financial time series are investigated for characteristics that may be used for volatility modelling. The well-known GARCH models as well as some related models are investigated, and their abilities to capture the characteristics of financial returns are evaluated in a large estimation study where the traditional estimation methods as well as an alternative recursive estimation method are used.
Parameters of bivariate Stochastic Volatility Models are traditionally very difficult to estimate. This thesis investigates an indirect estimation procedure, were the parameters of continuous time models are obtained from discrete time model parameter estimates. A new discrete time stochastic volatility model, SSSH-GARCH, is introduced.