Wavelet Basis Selection for Regression by Cross-Validation

  • Seth A. Greenblatt
Part of the Advances in Computational Economics book series (AICE, volume 6)

Abstract

Much of the attention paid to wavelets in the statistical literature has been in the area of nonparametric regression, smoothing, and denoising. Generally, the wavelet-based techniques involve finding an appropriate threshold level and using that threshold to either shrink or zero out wavelet coefficients, depending upon their sizes. Recently, a new technique for threshold determination based upon cross-validation has been introduced. In this study, we look at several criteria that may be used, in conjunction with cross-validation, to select the most appropriate wavelet basis for the problem at hand and then we apply the resulting technique to financial market data.

Keywords

Discrete Wavelet Transform Wavelet Packet Wavelet Basis Nonparametric Regression Chirp Signal 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. Barnett, V. and T. Lewis, 1994, Outliers in Statistical Data ( 3rd ed. ), Wiley Series in Probability and Mathematical Statistics, Chichester: Wiley.Google Scholar
  2. Coifman, R.R. and M.V. Wickerhauser, 1992, ‘Entropy based methods for best basis selection’, IEEE Trans. on Inf. Theory 38 (2), 719–746.CrossRefGoogle Scholar
  3. Daubechies, I., 1988, ‘Orthonormal bases of compactly supported wavelets’, Comm. Pure and Appl. Math. 41, 909–996.CrossRefGoogle Scholar
  4. Daubechies, I., 1992, Ten Lectures on Wavelets, Number 61 in CBMS-NSF Series in Applied Mathematics, Philadelphia: SIAM.Google Scholar
  5. Daubechies, I., 1993, ‘Wavelet transforms and orthonormal wavelet bases’, in I. Daubechies (Ed.), Different Perspectives on Wavelets, Volume 47 of Proceedings of Symposia in Applied Mathematics, pp. 1–33.Google Scholar
  6. Donoho, D.L., 1993, ‘Unconditional bases are optimal bases for data compression and for statistical estimation’, Appl. Comput. Harmon. Anal. 1, 100–115.CrossRefGoogle Scholar
  7. Donoho, D.L. and I.M. Johnstone, 1993, ‘Ideal spatial adaptation by wavelet shrinkage’, Preprint, Department of Statistics, Stanford University.Google Scholar
  8. Donoho, D.L. and I.M. Johnstone, 1994a, ‘Adapting to unknown smoothness by waveletGoogle Scholar
  9. shrinkage’, Journal of the American Statistical Association,accepted for publication. Donoho, D.L. and I.M. Johnstone, 1994b, ‘Ideal denoising in an orthonormal basis chosenGoogle Scholar
  10. from a library of bases’, Preprint, Department of Statistics, Stanford University.Google Scholar
  11. Donoho, D.L., I.M. Johnstone, G. Kerkyacharian, and D. Picard, 1994, ‘Wavelet shrinkage: Asymptopia?’, Journal of the Royal Statistical Society, B, to appear.Google Scholar
  12. Greenblatt, S.A., 1994, ‘Wavelets in econometrics’, Maple Technical Newsletter.Google Scholar
  13. Greenblatt, S.A., 1995, ‘Wavelets in econometrics: An application to outlier testing’, in M. Gilli (Ed.), Computational Economic Systems: Models, Methods, and Econometrics, Volume 5 of Advances in Computational Economics, Dordrecht, Kluwer Academic Publishers, pp. 139–160.Google Scholar
  14. Nason, G.P., 1994, ‘Wavelet regression by cross-validation’, Preprint, Department of Mathematics, University of Bristol.Google Scholar
  15. Taswell, C., 1995, ‘Near-best basis selection algorithms with non-additive information cost functions’, in Proc. IEEE-SP Symposium on TFTSA-94, IEEE-SP, to appear.Google Scholar
  16. Wickerhauser, M.V., 1991, ‘INRIA lectures on wavelet packet algorithms’, Lecture Notes, INRIA.Google Scholar
  17. Wickerhauser, M.V., 1993, ‘Best-adapted wavelet packet bases’, in Volume 47 of Proceedings of Symposia in Applied Mathematics, pp. 155–171.Google Scholar

Copyright information

© Springer Science+Business Media Dordrecht 1997

Authors and Affiliations

  • Seth A. Greenblatt

There are no affiliations available

Personalised recommendations