Activity Number:
|
65
|
Type:
|
Topic Contributed
|
Date/Time:
|
Monday, August 12, 2002 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Business & Economics Statistics Section*
|
Abstract - #300728 |
Title:
|
Modeling Nonstationary Time Series using Link Functions
|
Author(s):
|
Sujit Ghosh*+ and Dazhe Wang
|
Affiliation(s):
|
North Carolina State University and North Carolina State Unviersity
|
Address:
|
209F Patterson Hall, Raleigh, North Carolina, 27695-8203, USA
|
Keywords:
|
Bayesian Inference ; Link functions ; MCMC ; Nonstationary ; Time series ; Unit root
|
Abstract:
|
Many financial and economic time series exhibit nonstationarity and stochastic volatility. Modeling nonstationarity in such series has been the subject of a great debate for several decades. In general, nonstationarity of a time series occurs through trend, or due to the presence of a unit root, or both. One common approach is to detrend the model and then test the unit root hypothesis. We propose a new hierarchical Bayesian approach using link functions to model the structural changes in the stationarity parameter of the stochastic process, which in turn also affects the conditional mean of the process. The proposed models will allow us to test the nonstationarity of the time series without first detrending the model. In other words, our approach provides a unified framework to model the stationarity parameter and the trend simultaneously. We present some simulation studies to illustrate our proposed models.
|