Volume 1, Issue 5, December 2013, Page: 38-46
A New Method for the Model Selection in B-Spline Surface Approximation with an Influence Function
Hongmei Bao, Graduate School of Environmental Science, Okayama University, Okayama, Japan
Kaoru Fueda, Graduate School of Environmental and Life Science, Okayama University, Okayama, Japan
Received: Aug. 27, 2013;       Published: Oct. 20, 2013
DOI: 10.11648/j.sjams.20130105.11      View  2628      Downloads  124
Abstract
In model selection, the most effective method requires much time.The analysis of the bivariate B-spline model with a penalized term has many difficulties.It has many factors and parameters such the number of the knots, the locations of those knots, number of B-spline functions and the value of the smoothing parameter of the penalized term.For the determination of the model we have to compare a large amount of the combinations of those parameters. Various information criteria are considered and the cross validation (CV) criterion is excellent but it requires a large amount of computational costs. The effect of the influence function and the techniques of the generalized cross validation (CV) are considered. The influence function is related to the first term of a Taylor expansion. Some alternative methods are tested and a new method is proposed. For the verification of this method theoretical proof and the computational results are shown.
Keywords
B-Spline Surface, Generalized Information Criterion, Influence Function, Generalized Cross-Validation, Cross-Validation, Kullback-Leibler Divergence, Surface Model Selection
To cite this article
Hongmei Bao, Kaoru Fueda, A New Method for the Model Selection in B-Spline Surface Approximation with an Influence Function, Science Journal of Applied Mathematics and Statistics. Vol. 1, No. 5, 2013, pp. 38-46. doi: 10.11648/j.sjams.20130105.11
Reference
[1]
Konishi, S., Kitagawa, G. (1996). "Generalised information criteria in model selection.",Biometrika, 83, 875–890.
[2]
Stone,M., "Cross-validatory choice and assessment of statistical predictions (with discussion)", Journal of the Royal Statistical Society, Series B, 36 (1974), 111–147.
[3]
Ueki, M. and Fueda, K. (2010)."Optimal Tuning Parameter Estimaiton In Maximum Penalized Likelihood Method", Annals of the Institute of Statistical Mathematics, 62, 413-438.
[4]
Konishi, S., Kitagawa, G. (2008). "Information Criteria and Statistical Modeling", Springer Science+Business Media, LLC.
[5]
Good, I. J. and Gaskins, R.A.(1971). "Non parametric roughness penalties for probability densities", Biometrika, Vol. 58. pp. 255-277.
[6]
Good, I. J. and Gaskins, R.A.(1980). "Density estimation and bump hunting by the penalized likelihood method exemplified by scattering and meteorite data", Journal of American Standard Association, Vol. 75. pp. 42-56.
[7]
Green, P. J., Silverman, B. W.(1994). "Nonparametric Regression and Generalized Linear Models", Chapman and Hall, London.
[8]
Umeyama, S. (1996). "Discontinuity extraction in regularization using robust statistics", Technical report of IEICE.,PRU95-217 (1996). pp. 9-16.
[9]
Cox, M.G.(1972). "The numerical evaluation of B-splines", J. Inst. Math. Appl., 10, pp.134-149.
[10]
Cox, M.G.(1975). "An algorithm for spline interpolation", J. Inst. Math. Appl.,15, pp.95-108.
[11]
de Boor, C.(1972). "On calculation with B-splines", J. Approx. Theory, 6, pp.50-62.
[12]
Schoenberg, I. J., Whitney, A.(1953). "On Pólya frequency functions III", Trans. Amer. Math. Soc, Vol. 74. pp. 246-259, pp. 246-259.
[13]
Ueki, M. and Fueda, K.(2006)."Over close model problem and a modification of information criteria", Proceedings of The 28th symposium of Japanese Society of Applied Statistics,pp.7-10. (in Japanese)
Browse journals by subject