|
Activity Number:
|
88
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, July 30, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Business and Economics Statistics Section
|
| Abstract - #307822 |
|
Title:
|
Bayesian Change Point Model for Time Series
|
|
Author(s):
|
Sinsup Cho*+ and Juwon Kim and Seungmin Nam
|
|
Companies:
|
Seoul National University and Seoul National University and Samsung Fire & Marine Insurance Co., LTD
|
|
Address:
|
San 56-1, Shillim-Dong, Gwanak-Gu, Seoul, International, 151-742, Korea
|
|
Keywords:
|
Bayesian ; change point ; MCMC ; time series ; ARCH ; long memory
|
|
Abstract:
|
Change point detection in time series by Bayesian method is presented. The occurrence of the change point is modeled as the discrete time Markov process with unknown transition probabilities and is estimated by MCMC based on Chib's (1998) approach. The model assumes that all or part of the parameters in the change point model change over time and the time of the change points are known. We apply the algorithm to ARCH and ARFIMA models. Simulation is performed using a variant of perfect sampling algorithm to achieve the accuracy and efficiency. We compare the performance of the proposed change point model with the Kokoszka and Leipus (2000) CUSUM type estimator using AR(1)example. The yearly Nile-river data is analyzed as an example of the long memory process.
|
