To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
A Generalized Pseudo-Bayesian EM Algorithm for Switching Dynamic Factor Analysis |
|
Author |
Mohamed Saidane, Christian Lavergne |
| Citation |
Vol. 6 No. 5 pp. 30-37 |
|
Abstract |
In this article, we investigate a class of switching latent dynamic factor models for conditionally heteroscedastic financial time series. The proposed innovation in this paper can be viewed in two different ways. It can be presented as a generalization of the standard factor analysed hidden Markov model [1] since we will allow the common or specific variance of the model to be a stochastic function of time. Alternatively, we can present the contribution as an extension to conditionally heteroscedastic factor analysis [2] where the low dimensional subspace is modeled with a Gaussian hidden Markov model (HMM). For maximum likelihood estimation, we propose an iterative expectation-maximization (EM) algorithm based on a quasi-optimal switching Kalman filter approach combined with a generalized pseudo-bayesian approximation (GPB). The various regimes, the common factors and their volatilities are supposed unobservable and the inference must be carried out from the observable process. Extensive Monte Carlo simulations show promising results of the algorithms, especially for segmentation and tracking tasks. |
|
Keywords |
Factor Analysis, HMM, Conditional Heteroscedasticity, EM Algorithm, GPB Approximation |
|
URL |
