The Hessian Method (Highly Efficient State Smoothing, In a Nutshell)
I introduce the HESSIAN method for semi-Gaussian state space models with univariate states. The vector of states a=(a^1; ... ; a^n) is Gaussian and the observed vector y= (y^1 ; ... ; y^n )> need not be. I describe a close approximation g(a) to the density f(a|y). It is easy and fast to evaluate g(a) and draw from the approximate distribution. In particular, no simulation is required to approximate normalization constants. Applications include likelihood approximation using importance sampling and posterior simulation using Markov chain Monte Carlo (MCMC). HESSIAN is an acronym but it also refers to the Hessian of log f(a|y), which gures prominently in the derivation. I compute my approximation for a basic stochastic volatility model and compare it with the multivariate Gaussian approximation described in Durbin and Koopman (1997) and Shephard and Pitt (1997). For a wide range of plausible parameter values, I estimate the variance of log f(a|y) - log g(a) with respect to the approximate density g(a). For my approximation, this variance ranges from 330 to 39000 times smaller.
|Date of creation:||2008|
|Contact details of provider:|| Postal: CP 6128, Succ. Centre-Ville, Montréal, Québec, H3C 3J7|
Phone: (514) 343-6540
Fax: (514) 343-5831
Web page: http://www.sceco.umontreal.ca
More information through EDIRC
Please report citation or reference errors to , or , if you are the registered author of the cited work, log in to your RePEc Author Service profile, click on "citations" and make appropriate adjustments.:
- Geweke, John, 1989. "Bayesian Inference in Econometric Models Using Monte Carlo Integration," Econometrica, Econometric Society, vol. 57(6), pages 1317-1339, November.
- Jacquier, Eric & Polson, Nicholas G & Rossi, Peter E, 2002.
"Bayesian Analysis of Stochastic Volatility Models,"
Journal of Business & Economic Statistics,
American Statistical Association, vol. 20(1), pages 69-87, January.
- Jacquier, Eric & Polson, Nicholas G & Rossi, Peter E, 1994. "Bayesian Analysis of Stochastic Volatility Models," Journal of Business & Economic Statistics, American Statistical Association, vol. 12(4), pages 371-389, October.
- McCAUSLAND, William J. & MILLER, Shirley & PELLETIER, Denis, 2007. "A New Approach to Drawing States in State Space Models," Cahiers de recherche 07-2007, Centre interuniversitaire de recherche en économie quantitative, CIREQ.
- William J. McCausland & Shirley Miller & Denis Pelletier, 2007. "A New Approach to Drawing States in State Space Models," Working Paper Series 014, North Carolina State University, Department of Economics, revised Aug 2007.
- McCAUSLAND, William J. & MILLER, Shirley & PELLETIER, Denis, 2007. "A New Approach to Drawing States in State Space Models," Cahiers de recherche 2007-06, Universite de Montreal, Departement de sciences economiques.
- Liesenfeld, Roman & Richard, Jean-Francois, 2003. "Univariate and multivariate stochastic volatility models: estimation and diagnostics," Journal of Empirical Finance, Elsevier, vol. 10(4), pages 505-531, September.
- J. Durbin, 2002. "A simple and efficient simulation smoother for state space time series analysis," Biometrika, Biometrika Trust, vol. 89(3), pages 603-616, August.
- Sangjoon Kim & Neil Shephard & Siddhartha Chib, 1998. "Stochastic Volatility: Likelihood Inference and Comparison with ARCH Models," Review of Economic Studies, Oxford University Press, vol. 65(3), pages 361-393.
- repec:bla:restud:v:65:y:1998:i:3:p:361-93 is not listed on IDEAS
- Håvard Rue & Ingelin Steinsland & Sveinung Erland, 2004. "Approximating hidden Gaussian Markov random fields," Journal of the Royal Statistical Society Series B, Royal Statistical Society, vol. 66(4), pages 877-892.
- Richard, Jean-Francois & Zhang, Wei, 2007. "Efficient high-dimensional importance sampling," Journal of Econometrics, Elsevier, vol. 141(2), pages 1385-1411, December. Full references (including those not matched with items on IDEAS)