Date(s) : 10/04/2017 iCal
15 h 30 min - 16 h 30 min
We introduce a general state-space (or latent factor) model for time series and panel data. The state process has a polynomial expansion based dynamics that can approximate any Markov dynamics arbitrarily well, and has a latent, endogenous switching regime interpretation. The resulting state-space model is associated with simulation-free, recursive formulas for prediction and filtering, as well as the maximum composite likelihood estimation method, which has an extremely low computational cost. When applied to the stochastic volatility (SV) of asset returns, the model captures, in a unified framework, stylized facts such as heavy tailed return, volatility feedback, as well as time irreversibility. The methodology is illustrated using Apple stock return data, which confirms the improvement of our model with respect to a benchmark SV model.
https://www.greqam.fr/en/users/lu
Catégories Pas de Catégories