A Bayesian Simulation Approach to Inference on a Multi-State Latent Factor Intensity Model
Melbourne Institute Working Paper No. 16/08
Date: August 2008
This paper provides a Bayesian approach to inference on a multi-state latent factor intensity model to manage the problem of highly analytically intractable pdfs. The sampling algorithm used to obtain posterior distributions of the model parameters includes a particle filter step and a Metropolis-Hastings step within a Gibbs sampler. A simulated example is conducted to show the feasibility and accuracy of this sampling algorithm. The approach is applied to the case of credit ratings transition matrices.