Skip to content
kboudt edited this page Mar 17, 2016 · 1 revision

Background

Modeling the univariate and multivariate conditional distribution of relevant financial and economic variables is of primary importance for understanding complicated real-world problems. Examples are the spillover between returns of financial assets and important state variables such as the VIX and TED spread.

Recently, a new class of observation driven models (Cox et al., 1981), named Generalised Autoregressive Score (GAS) models, has been proposed by Creal et al. (2013) and Harvey (2013). GAS models are extremely flexible parametric models particularly suited to update the conditional distribution of time-dependent random variables.

The main feature of GAS model is that the updating recursion of unobserved model parameters is based on the score of the conditional distribution of the data. This updating mechanism make GAS model nested with many of the well known observation driven model, like the normal GARCH model.

They can be easily specified for univariate and multivariate random variables and are found to be able to accurately approximate complicated non–linear non–Gaussian state space models for which the estimation procedure could be prohibitive (see, e.g. Koopman et al. 2015). The usual GAS models have been extended to include exogenous regressors (Salvatierra and Patton, 2015; Boudt et al., 2012), Markov Switching coefficients (Boudt et al., 2012; Bernardi and Catania, 2015), trend and seasonal components (Caivano et al., 2015) are also possible.

Related work

Even if the GAS literature have been growing fast during last three years, computer packages for the estimation of these kind of models are still rare. The only two available packages are the betategarch R package of Sucarrat (2015) and the DySco Ox package of Andres (2014). However, these two computer packages only cover particular univariate models and an integrated environment for GAS models is still not available.

Details of your coding project

The structure of the package should follow existing well-known/standard packages dealing with multivariate modeling such as the rmgarch package of Ghalanos (2015). It should provide users with several tools to estimate, simulate and test multivariate GAS model with the possibility to exploit the information in external variables.

Parametric distributions that can be included are: Gaussian, Student–t, Skew–Student–t, ALD, Multivariate Gaussian, Multivariate Student–t, Gaussian Copula, t–Copula.

The package could include one step ahead and multi-step ahead forecasting routines as well as confidence bands for filtered and predicted dynamics following the recent paper of Blasques et al. (2015)?

Extension of classical GAS models to include exogenous variable or Markow switching coefficients can be also considered.

The package should be structured using S4 objects. The main routines should be written in C++ using the Rcpp and RcppArmadillo packages of Eddelbuettel (2016a) and Eddelbuettel (2016b) for speedup purposes. Part of the code that can be parallelised should make use of openMP or the parallel package.

The output of the routines has to be clearly understandable and the graphical part of the package should also play an important role.

The project will be developed with https://www.rstudio.com/ and stored on https://github.com.

Expected impact

The R language has become an important vector for knowledge transfer in the multivariate analysis of economic and financial time series. Currently, there is no package dealing with the expanding field of multivariate GAS models. This package will therefore provide a new useful tool to practitioners and academics in the economic and financial community.

Mentors

Prof. Dr. Kris Boudt and Prof. Dr. David Ardia

Tests

Applicants have to be able to show that they have:

  • A good working knowledge of programming in R, Rcpp and C++.
  • A good working knowledge of Roxygen for the documentation.
  • A good working knowledge of knitr/LaTeX for the vignette.
  • Familiarities with the construction of R packages.
  • Good coding standards (Google’s C++ and R style guide).
  • Familiarities with GAStype model.
  • Familiarities with Maximum Likelihood estimation.

References

Andres, P. (2014). Maximum likelihood estimates for positive valued dynamic score models; the dysco package. Computational Statistics & Data Analysis, 76:34-42.

Bernardi, M. and Catania, L. (2015). Switching-GAS Copula Models With Application to Systemic Risk. ArXiv e-prints.

Blasques, F., Koopman, S. J., Lasak, K., and Lucas, A. (2015). In-sample confidence bands and out-of-sample forecast bands for time-varying parameters in observation driven models. International Journal of Forecasting, (forthcom- ing).

Boudt, K., Danielsson, J., Koopman, S. J., and Lucas, A. (2012). Regime switches in the volatility and correlation of financial institutions. National Bank of Belgium Working Paper, (227).

Caivano, M., Harvey, A., and Luati, A. (2015). Robust time series models with trend and seasonal components. SERIEs, pages 1-22.

Cox, D. R., Gudmundsson, G., Lindgren, G., Bondesson, L., Harsaae, E., Laake, P., Juselius, K., and Lauritzen, S. L. (1981). Statistical analysis of time series: Some recent developments [with discussion and reply]. Scandinavian Journal of Statistics, pages 93-115.

Creal, D., Koopman, S. J., and Lucas, A. (2013). Generalized autoregressive score models with applications. Journal of Applied Econometrics, 28(5):777-795.

Eddelbuettel, D., Francois, R., Allaire, J., Ushey, K., Kou, Q., Bates, D., and Chambers, J. (2016a). rcpp: Seamless r and c++ integration. r package version 0.12.3 . url: https://cran.r-project.org/web/packages/rcpp/index.html.

Eddelbuettel, D., Francois, R., and Bates, D. (2016b). rcpparmadillo: ‘rcpp’ in- tegration for the ‘armadillo’ templated linear algebra library. r package version 3.2. url: https://cran.r-project.org/web/packages/rcpparmadillo/index.html.

Harvey, A. C. (2013). Dynamic Models for Volatility and Heavy Tails: With Applications to Financial and Economic Time Series. Cambridge University Press.

Koopman, S. J., Lucas, A., and Scharth, M. (2015). Predicting time-varying parameters with parameter-driven and observation-driven models. Review of Economics and Statistics, forthcoming.

Salvatierra, I. D. L. and Patton, A. J. (2015). Dynamic copula models and high frequency data. Journal of Empirical Finance, 30:120-135.

Sucarrat, G. (2015). betategarch: Simulation, estimation and forecasting of beta-skew-t-egarch models. r package version 3.2. url: https://cran.r-project.org/web/packages/betategarch/.

Clone this wiki locally