A1299
Title: State-dependent copula particle filtering for stochastic volatility models
Authors: Tingfu Chen - National Central University (Taiwan) [presenting]
Abstract: A state-dependent copula-particle filter (SD-CoPF) is developed for the joint estimation of parameters and latent states in multivariate stochastic volatility models. By treating model parameters as time-varying states and incorporating kernel smoothing, the framework ensures convergence of posterior distributions and mitigates particle degeneracy. The methodology integrates copula functions to capture complex cross-market dependency structures, distinguishing between particle-correlated and state-correlated filtering architectures. Theoretical properties, including parameter convergence and the validity of the auxiliary-like importance weighting scheme, are mathematically established. Numerical simulations across Gaussian, Student-t, and Archimedean copula families demonstrate that SD-CoPF achieves high estimation accuracy. The proposed algorithm effectively tracks hidden volatility dynamics and adapts to various non-linear dependence structures in high-dimensional financial time series.
