Parameter Estimations of Stochastic Volatility Model by Modified Adaptive Kalman Filter with QML

To determine the parameters of Stochastic Volatility Model (SVM), a modification to the Quasi Maximum Likelihood (QML) scheme has been proposed by employing (modified) Adaptive Kalman Filter (AKF). AKF allows optimization over lesser number of parameters as the variance (sigma(2)(v)) of the noise in the volatility state equation is determined by the AKF. The adaptive method, instead of a constant sigma(2)(v), allows it to be time varying. Before applying the methodology on market data, the proposed method is characterized here by synthetic data through simulation investigations. Numerical experiments show that the performance of SVM based QML-KF and novel QML-AKF are comparable to that of more popular GARCH family based techniques.