Odhad vybraných typů modelů finančních aktiv

Abstract

Diploma thesis is devoted to the estimation of selected types of models of financial assets. Its aim is to estimate a selection of the best models for various financial assets with varying frequency data. The selected financial instruments are Google shares and stock market index S&P500 with a daily, weekly, monthly frequency and rate of the euro with a daily frequency data. Test normality of the probability distribution of yields was made for these financial time series and calculated some characteristics of location and variability. Using the method of maximum likelihood estimate of selected models of volatility GARCH and EGARCH assuming normality was made and it was examined which model better reflects reality. Criterion of evaluation in selecting the best model was the size of maximizing the likelihood function, the smaller the standard deviation of a graphical comparison of actual and model, the better model has greater short-term volatility. Models GARCH and EGARCH were not for modeling yields very accurate, almost always modeling yields fluctuated only slightly around zero and failed to capture significant fluctuations in real yields. For dispersion modeling yields of those financial instruments with different frequencies EGARCH model respond better the reality in all cases. It was found that asymmetric EGARCH model is for the modeling of financial time series much wider use than the symmetric GARCH model.