Gholamhossein Gholami - CEREMADE Universite Paris-Dauphine France

Bayesian change-point problem deals with sudden change in the distribution of the given data. A relevant case is change-point in Stochastic Volatility modeling. The SV models deal with time-varying volatility. These models are based on two processes, the volatility (hidden) and the innovation. Studying the behavior of the hidden process, in term of changing the parameters over time, is of interest. In this work, considering a uniform prior for the change-point and conjugate priors for other parameters, we estimate the model. As the posterior distribution is complex and not tractable, MCMC methods, particularly Gibbs and Mertopolis-Hastings algorithms, are used.

Bayesian change-point problem; Stochastic Volatility models; Monte Carlo Markov Chain Methods (MCMC); Gibbs and Metropolis-Hastings algorithm.