Modelling of Stock Returns Time-Series

In the study submitted, selected methods of financial time-series analysis are applied to daily returns of the most liquid stocks at Czech capital market. In most cases, symmetric GARCH(1,1) models are quite satisfactory. Further, ARFIMA models enabling to catch "long memory" of underlying processes are suitable for the modelling both absolute values of returns and their volatility. Alternative posibility is to employ bilinear models, which prove to be suitable namely for returns. Hurst exponents computed signalize some tendency to cycles creation in some cases (ČEZ, Unipetrol).