Generalized Autoregressive Score models
GAS estimated volatility paths for Nordpool electricity prices based on the Student's t
distribution and the Gaussian distribution. The Gaussian GAS volatility model coincides with the familiar GARCH model (more information
Generalized Autoregressive Score (GAS) models, also known as Dynamic Conditional Score (DCS) models, provide a general framework for modeling time variation in parametric models.
The key features are:
These models have been applied successfully in areas such as default and credit risk modeling, stock volatility and correlation modeling, modeling time varying dependence structures, CDS spread modeling and questions relating to financial stability and systemic risk, modeling high frequency data, etc.
- easy estimation and inference: the likelihood is available in closed form;
- generality: you are in business whenever you can compute the score of your parametric conditional observation density with respect to the time varying parameter.
On this site you find:
- some background information about GAS models;
- the papers about GAS and score driven models that we
are aware of;
- computer code for GAS models to help you get started;
- Feb 28, 2018: Added the example code in the codes section for
"Semiparametric score driven volatility models".
- Feb 19, 2018: Added the example code in the codes section for "Bank business models at zero interest rates": clustering with time varying means and covariances.
- Feb 5, 2018: Added new entries to the papers section: Chen; Buccheri; Lit et al.; Hansen and Schmidthblaicher; and updated published and forthcoming entries.
- Feb 2, 2018: The main Generalized Autoregressive Score paper is in the TOP 10 of most downloaded published articles of all times of the Journal of Applied Econometrics. Really happy! See the journal newsletter.
- Nov 15, 2017: added a paper to the papers section: Lucas, Schaumburg, Schwaab (in press, JBES).
- Nov 14, 2017: added 2 updates to the papers section: link correction and Harvey and Lange (2017) appeared.
- Oct 7, 2017: added 15 new entry and updates to the papers section after some googling over 2016 and 2017.
- Sep 6, 2017: added three new entry in the papers section: modeling VaR and ES by scores (Patton et al., 2017), portfolio selection based on scores (Bernardi and Catania, 2016), and extreme values modeling using scores (Chong et al., 2017).
- Sep 4, 2017: added a new entry in the papers section on modeling time varying tail shapes: Massacci (in press).
- Sep 4, 2017: added a new entry in the papers section on modeling crypto-currencies with GAS models: Catania and Grassi (2017).
- Sep 4, 2017: updated the GAS HMM package of Marco Bazzi in the code section.
- Jul 6, 2017: added two new entries in the papers section on accellerated GAS and on GAS for football matches and GAS for option pricing: Koopman and Lit (2017) and Blasques, Gorgi, Koopman (2017) and Lin and Yu (2017).
- Jul 4, 2017: David Kranenburg and Rutger Lit improved the example matlab code for the GAS volatility model (considerably faster without globals).
- May 31, 2017: added three new entries in the papers section by Blazsek et al. 2016, 2017): time varying volatility, regime switching, and time varying degrees of freedom.
- May 16, 2017: updated and added entries in the papers section, including Neves et al. on GAS for actuarial models.
- Feb 16, 2017: new entries in the papers section: added the Oh and Patton paper forthcoming in JBES.
- Feb 9, 2017: new entries in the papers section: working paper of Buccheri et al. (2017) on GAS for high-frequency and synchronicity.
- Feb 6, 2017: two new entries in the papers section: Billé and Catania forthcoming in Journal of Applied Econometrics and working paper by Catania and Nonejad.
Send us your feedback or your contributions on GAS modeling at firstname.lastname@example.org.
Before using any of the software, please read the legal disclaimer.