Advances and Applications in Statistics
Volume 37, Issue 2, Pages 149 - 170
(December 2013)
|
|
BAYESIAN WHITTLE ESTIMATION OF ARFIMA MODEL
Haruhisa Nishino and Kazuhiko Kakamu
|
Abstract: In this paper, we propose a Bayesian estimation method of autoregressive fractionally integrated moving average (ARFIMA) models using the Markov chain Monte Carlo (MCMC) method, where a Whittle likelihood is used instead of a time domain one. A normal approximation using a second-order Taylor expansion enables us to implement the MCMC. The proposed estimation method is supported by Monte Carlo experiments and applied to the realized volatility data of the NIKKEI 225 and TOPIX indices. |
Keywords and phrases: long memory, ARFIMA model, Whittle estimation, Bayesian approach, Markov chain Monte Carlo (MCMC) method, realized volatility. |
|
Number of Downloads: 355 | Number of Views: 1113 |
|