Vlastnosti ML estimátorů konečných délek časových řad ve score-driven volatility modelech

The aim of this thesis is to propose a simulation study on the selected Generalized Autoregressive Score (GAS) models and assess the properties of the ML estimators depending on different lengths of time series. Examined are the most important and widely used GAS models (two with normal distribution and one with Student's t-distribution). Each is using variance as a time-varying parameter, which can be interpreted as volatility of the time series. The simulation study also investigates the impact of the choice of the optimization algorithm in the parameters estimation. Another dimension considered is the actual setup of the data-generating processes and the associated level of volatility persistence in the given time series.