|
Activity Number:
|
294
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, August 5, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Business and Economics Statistics Section
|
| Abstract - #300255 |
|
Title:
|
An Empirical Comparison of Some Parameter Estimation Methods in Stochastic Volatility Models
|
|
Author(s):
|
Bovas Abraham*+ and Ji Eun Choi
|
|
Companies:
|
University of Waterloo and University of Waterloo
|
|
Address:
|
Dept. of Statistics and Act. Sci, Waterloo, ON, N2L 3G1, Canada
|
|
Keywords:
|
Stochastic Volatility ; Simulated Maximum Likelihood ; Markov Chain Monte Carlo
|
|
Abstract:
|
Financial time series often exhibit time-dependent variances (volatility clustering) and excess kurtosis in the marginal distributions. One class of models which captures those features is the stochastic volatility (SV) models. In these models, time-dependent variances are assumed to be random variables generated by an underlying latent stochastic process. A standard SV model assumes that the conditional distribution of observations is normal and the volatility sequence evolves as an autoregressive sequence with log normal marginals. Exact maximum likelihood estimation is difficult in the SV models and several approximate methods are proposed in the literature. In this paper we study the Simulated Maximum Likelihood (SML) and Markov Chain Monte Carlo (MCMC) methods for estimating the parameters of a standard SV model.
|
