抄録
Nonlinear non-Gaussian state space model is considered for the analysis of nonstationary time series. Recursive formulas for state estimation such as predictor, filter and smoother can be derived for this model. These formulas can be realized by using Monte Carlo filter and smoother. These recursive computational methods facilitates very flexible nonlinear non-Gaussian modeling of nonstationary time series. As examples of this method, two types of modeling of volatility of financial time series are shown.
