A graphical tool for selecting the number of slices and the dimension of the model in SIR and SAVE approaches

Computational Statistics volume 27, pages103–125(2012)Cite this article

Abstract

Sliced inverse regression (SIR) and related methods were introduced in order to reduce the dimensionality of regression problems. In general semiparametric regression framework, these methods determine linear combinations of a set of explanatory variables X related to the response variable Y, without losing information on the conditional distribution of Y given X. They are based on a “slicing step” in the population and sample versions. They are sensitive to the choice of the number H of slices, and this is particularly true for SIR-II and SAVE methods. At the moment there are no theoretical results nor practical techniques which allows the user to choose an appropriate number of slices. In this paper, we propose an approach based on the quality of the estimation of the effective dimension reduction (EDR) space: the square trace correlation between the true EDR space and its estimate can be used as goodness of estimation. We introduce a naïve bootstrap estimation of the square trace correlation criterion to allow selection of an “optimal” number of slices. Moreover, this criterion can also simultaneously select the corresponding suitable dimension K (number of the linear combination of X). From a practical point of view, the choice of these two parameters H and K is essential. We propose a 3D-graphical tool, implemented in R, which can be useful to select the suitable couple (H, K). An R package named “edrGraphicalTools” has been developed. In this article, we focus on the SIR-I, SIR-II and SAVE methods. Moreover the proposed criterion can be use to determine which method seems to be efficient to recover the EDR space, that is the structure between Y and X. We indicate how the proposed criterion can be used in practice. A simulation study is performed to illustrate the behavior of this approach and the need for selecting properly the number H of slices and the dimension K. A short real-data example is also provided.

This is a preview of subscription content, log in to check access.

References

  1. Aragon Y, Saracco J (1997) Sliced inverse regression (SIR): an appraisal of small sample alternatives to slicing. Comput Stat 12: 109–130

    MathSciNet  Google Scholar 

  2. Breiman L, Friedman JH (1985) Estimating optimal transformations for multiple regression and correlation (with discussion). J Am Stat Assoc 80: 580–619

    MathSciNet  MATH  Article  Google Scholar 

  3. Carroll RJ, Li KC (1992) Measurement error regression with unknown link: dimension reduction and data visualization. J Am Stat Assoc 87: 1040–1050

    MathSciNet  MATH  Article  Google Scholar 

  4. Chen H (1991) Estimation of a projection-pursuit type regression model. Ann Stat 19: 142–157

    MATH  Article  Google Scholar 

  5. Chen CH, Li KC (1998) Can SIR be as popular as multiple linear regression?. Stat Sin 8: 289–316

    MATH  Google Scholar 

  6. Cook RD (2000) SAVE: a method for dimension reduction and graphics in regression. Commun Stat Theor Methods 29: 2109–2121

    MATH  Article  Google Scholar 

  7. Cook RD, Li B (2002) Dimension reduction for conditional mean in regression. Ann Stat 30: 455–474

    MathSciNet  MATH  Article  Google Scholar 

  8. Cook RD, Weisberg S (1991) Discussion of “Sliced inverse regression”. J Am Stat Assoc 86: 328–332

    Article  Google Scholar 

  9. Cook RD (1998) Regression graphics. Ideas for studying regressions through graphics. Wiley Series in Probability and Statistics. Wiley, New York

    Google Scholar 

  10. Cook RD, Weisberg S (1999) Applied statistics including computing and graphics. Wiley, New York

    Google Scholar 

  11. Duan N, Li KC (1991) Slicing regression: a link-free regression method. Ann Stat 19: 505–530

    MathSciNet  MATH  Article  Google Scholar 

  12. Efron B (1982) The jackknife, the bootstrap and other resampling plans. CBMS-NSF regional conference series in applied mathematics, 38. Society for Industrial and Applied Mathematics (SIAM), Philadelphia

  13. Ferré L (1997) Dimension choice for sliced inverse regression based on ranks. Student 2: 95–108

    Google Scholar 

  14. Ferré L (1998) Determining the dimension in sliced inverse regression and related methods. J Am Stat Assoc 93: 132–140

    MATH  Article  Google Scholar 

  15. Friedman JH, Stuetzle W (1981) Projection pursuit regression. J Am Stat Assoc 76: 817–823

    MathSciNet  Article  Google Scholar 

  16. Gannoun A, Saracco J (2003) An asymptotic theory for SIR α method. Stat Sin 13: 297–310

    MathSciNet  MATH  Google Scholar 

  17. Hall P (1989) On projection pursuit regression. Ann Stat 17: 573–588

    MATH  Article  Google Scholar 

  18. Hall P, Li KC (1993) On almost linearity of low dimensional projections from high dimensional data. Ann Stat 21: 867–889

    MathSciNet  MATH  Article  Google Scholar 

  19. Hsing T (1999) Nearest neighbor inverse regression. Ann Stat 27: 697–731

    MathSciNet  MATH  Article  Google Scholar 

  20. Hsing T, Carroll RJ (1992) An asymptotic theory for sliced inverse regression. Ann Stat 20: 1040–1061

    MathSciNet  MATH  Article  Google Scholar 

  21. Kötter T (2000) Sliced inverse regression. In: Schimek MG (ed) Approaches, computation, and application. Wiley, New York, pp 497–512

    Google Scholar 

  22. Li KC (1991) Sliced inverse regression for dimension reduction, with discussion. J Am Stat Assoc 86: 316–342

    MATH  Article  Google Scholar 

  23. Li KC (1992) On principal Hessian directions for data visualization and dimension reduction: another application of Stein’s lemma. J Am Stat Assoc 87: 1025–1039

    MATH  Article  Google Scholar 

  24. Li Y, Zhu L-X (2007) Asymptotics for sliced average variance estimation. Ann Stat 35: 41–69

    Article  Google Scholar 

  25. Liquet B, Saracco J (2008) Application of the bootstrap approach to the choice of dimension and the α parameter in the SIR α method. Commun Stat Simul Comput 37: 1198–1218

    MathSciNet  MATH  Article  Google Scholar 

  26. Prendergast LA (2007) Implications of influence function analysis for sliced inverse regression and sliced average variance estimation. Biometrika 94: 585–601

    MathSciNet  MATH  Article  Google Scholar 

  27. Saracco J (1997) An asymptotic theory for sliced inverse regression. Commun Stat Theor Methods 26: 2141–2171

    MathSciNet  MATH  Article  Google Scholar 

  28. Saracco J (2001) Pooled slicing methods versus slicing methods. Commun Stat Simul Comput 30: 489–511

    MathSciNet  MATH  Article  Google Scholar 

  29. Xia Y, Tong H, Li WK, Zhu L-X (2002) An adaptative estimation of dimension reduction space. J R Stat Soc B 64: 363–410

    MathSciNet  MATH  Article  Google Scholar 

  30. Ye Z, Weiss RE (2003) Using the bootstrap to select one of a new class of dimension reduction methods. (English summary). J Am Stat Assoc 98: 968–979

    MathSciNet  MATH  Article  Google Scholar 

  31. Yin X, Seymour L (2005) Asymptotic distributions for dimension reduction in the SIR-II method. Stat Sin 15: 1069–1079

    MathSciNet  MATH  Google Scholar 

  32. Zhu LX, Fang KT (1996) Asymptotics for kernel estimate of sliced inverse regression. Ann Stat 24: 1053–1068

    MathSciNet  MATH  Article  Google Scholar 

  33. Zhu L-X, Miao B, Peng H (2006) On sliced inverse regression with high dimensional covariates. J Am Stat Assoc 101: 630–643

    MathSciNet  MATH  Article  Google Scholar 

  34. Zhu L-X, Ohtaki M, Li YX (2007) On hybrid methods of inverse regression based algorithms. Comput Stat Data Anal 51: 2621–2635

    MathSciNet  MATH  Article  Google Scholar 

  35. Zhu LX, Ng KW (1995) Asymptotics of sliced inverse regression. Stat Sin 5: 727–736

    MathSciNet  MATH  Google Scholar 

  36. Zhu L-P, Zhu L-X (2007) On kernel method for sliced average variance estimation. J Multivar Anal 98: 970–991

    MATH  Article  Google Scholar 

  37. Zhou Y, Zhu L-P, Zhu L-X (2007) On splines approximations for sliced average variance estimation. J Stat Plan Inf (to appear)

Download references

Author information

Affiliations

  1. INSERM U897, ISPED, Université Victor Segalen Bordeaux 2, 146 rue Leo Saignat, 33076, Bordeaux Cedex, France

    Benoît Liquet

  2. INRIA Bordeaux Sud-Ouest, CQFD Team, Talence Cedex, France

    Jérôme Saracco

  3. Institut de Mathématiques de Bordeaux, UMR CNRS 5251, Université Bordeaux 1, 351 cours de la libération, 33405, Talence Cedex, France

    Jérôme Saracco

Authors
  1. Benoît Liquet
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Jérôme Saracco
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Jérôme Saracco.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Liquet, B., Saracco, J. A graphical tool for selecting the number of slices and the dimension of the model in SIR and SAVE approaches. Comput Stat 27, 103–125 (2012). https://doi.org/10.1007/s00180-011-0241-9

Download citation

Keywords

  • Bootstrap
  • Dimension reduction
  • Sliced inverse regression (SIR)
  • Sliced average variance estimation (SAVE)
  • Square trace correlation