05/09/2014
Ano: 2014
Código: WPE – 343
Autores/Pesquisadores:
Abstract:
State-space models are commonly used in the engineering, economic, and statisticalliteratures. They are exible and encompass many well-known statistical models, includingrandom coe cient autoregressive models and dynamic factor models. Bayesiananalysis of state-space models has attracted much interest in recent years. However,for large scale models, prior speci cation becomes a challenging issue in Bayesian inference.In this paper, we propose a exible prior for state-space models. The proposedprior is a mixture of four commonly entertained models, yet achieves parsimony inhigh-dimensional systems. Here parsimony” is represented by the idea that in a largesystem, some states may not be time-varying. Simulation and simple examples areused throughout to demonstrate the performance of the proposed prior. As an application,we consider the time-varying conditional covariance matrices of daily logreturns of 94 components of the S&P 100 index, leading to a state-space model with94 95/2=4,465 time-varying states. Our model for this large system enables us to useparallel computing.
 |
Download do arquivo no formato PDF |