No. 855

Title:

An innovation approach for the estimation, selection and prediction of discretely observed continuous-time stochastic volatility models

Author(s):

Tohru Ozaki (Institute of Statistical Mathematics) ;
J.C. Jimenez (Instituto de Cibernetica)

Key words:

stochastic volatility, innovation, method, Local Linearization filters, inference of diffusion processes

Abstract:

In this paper the estimation, selection and prediction of discretely observed continuous-time stochastic volatility models is carry out trough the innovation approach. The central idea consists in setting the stochastic volatility models inside the framework of nonlinear continuous-discrete state space models with multiplicative noise. Then, the unknown model parameters and the unobserved components of the stochastic volatility models are estimated by the approximated innovation method based on the Local Linearization filter. The Akaike Information Criteria and the prediction estimates obtained by the Local Linearization filter are used, respectively, for the model selection and for the one day head prediction of the market.