Computational Economics

, Volume 28, Issue 3, pp 277–290 | Cite as

Optimizing the Garch Model–An Application of Two Global and Two Local Search Methods

  • Kwami Adanu
Article

Abstract

Results from our optimization exercise clearly show the advantage of using the random search algorithms when we anticipate the search for the global optimum to be difficult. When the number of parameters in the model is relatively small (nine parameters) Differential Evolution performs better than Genetic Algorithm. However, when the number of parameters in the model is relatively large (fifteen parameters) the reverse case is true. A comparison of the Quasi-Newton and Simplex methods also shows that both the Quasi-Newton algorithm of shazam and the simplex algorithm of fminsearch are sensitive to starting values. However, allowing shazam to set its starting values or using the PRESAMP option to set the starting values produced the best results for shazam. The general conclusion of this paper is that the choice of optimization technique for difficult optimization problems like the one attempted here should be based on problem attributes. When in doubt, multiple techniques should be applied and the estimated results evaluated.

Keywords

GARCH global optimum genetic algorithm differential evolution Quasi-Newton algorithm Simplex method 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. Doornik, J.A. (2000). Multimodality and the GARCH likelihood. A Paper Presented at The World Conference of Econometric Society.Google Scholar
  2. Holland, J. (1992). Adaptation in Natural and Artificial Systems, 2nd edition. MIT Press.Google Scholar
  3. Houck, C., Joines, J. and Kay, M. (1996). A Genetic Algorithm for Function Optimization: A Matlab Implementation. ACM Transactions on Mathematical Software. Google Scholar
  4. Jerrel, M. (2000). Applications of Public Domain Global Optimization Software to Difficult Econometric Functions. Computing in Economics and Finance, No. 161.Google Scholar
  5. Lagarias, J.C., Reeds, J.A., Wright, M.H. and Wright, P.E. (1998). Convergence Properties of the Nelder-Mead Simplex Method in Low Dimensions. SIAM Journal of Optimization, 9(1), 112–147.CrossRefGoogle Scholar
  6. Storn, R. and Price, K. (1995). Differential Evolution – A Simple and Efficient Adaptive Scheme for Global Optimization Over Continuous Spaces. Technical Report TR-95-012, ICSI.Google Scholar
  7. The solver for the GA (GAOT.ZIP) written by Houck et al. (1996) can be found at the following HTML address: http://www.ie.ncsu.edu/mirage/GAToolBox/gaot/.
  8. The solver for the DE (Devec3.m) can also be found at the following internet address: http://www.icsi.berkeley.edu/storn/code.html#matl.

Copyright information

© Springer Science + Business Media, Inc. 2006

Authors and Affiliations

  • Kwami Adanu
    • 1
  1. 1.Michigan State UniversityEast LansingUSA

Personalised recommendations