Summary
Several methods are considered for the reconstruction of smooth functions and surfaces from noisy measurements. These methods are statistically motivated and belong to the area of nonparametric regression. They include the kernel method, local orthonormal polynomial expansions with prebinning, and locally weighted least squares. The close connections which exists between these methods are pointed out, and equivalent kernels for locally weighted least squares estimates are derived. An application to surface approximation for Iranian soil clay thickness data is discussed. We also consider the case of surfaces with discontinuities. This includes the edge estimation problem, maximum and cube splitting methods for edge estimation and associated two-step procedures for the recovery of the entire surface.
Similar content being viewed by others
References
I.A. Ahmad, P.E. LinFitting a multiple regression function. J. Statis. Plann. & Inf. 9 (1984), 163–176.
M.H. Alemi, A.S. Azari, D.R. NielsenKriging and univariate modeling of spatially correlated data. Soil Technology. 1 (1988), 133–147.
A.S. Azari, H.G. MüllerPreaveraged localized orthogonal polynomial estimators for surface smoothing and partial differentiation. J. Amer. Statist. Assoc 87 (1992), 1005–1017.
A.S. Azari, H.G. MüllerOrthogonal polynomial and hybrid estimators for nonparametric regression. J. Royal Statist. Soc. A, in press.
M.S. BartlettSmoothing periodograms from time series with continuous spectra. Nature 161 (1948), 486–487.
M.S. BartlettStatistical estimations for density functions. Sankhya A25 (1963), 245–254.
M.S. Bartlett, T. MedhiThe efficiency of procedures for smoothing periodograms from time series with continuous spectra. Biometrika 42 (1955), 143–150.
L. Breiman, J. FriedmanEstimating optimal transformations for multiple regression and correlation (with discussion). J. Amer. Statist. Assoc. 80 (1985), 580–619.
E. Carlstein, C. KrishbamoorthyBoundary estimations. J. Royal Statist. Soc., B39 (1992), 107–113.
R.M. ClarkNonparametric estimation of a smooth regression function. J. Royal Statist. Soc., B39 (1977), 107–113.
W. Cleveland, S. DevlinLocally weighted regression: An approach to regression analysis by local fitting. J. Amer. Statist. Assoc. 83 (1988), 569–610.
P.J. DaniellDiscussion of paper by M.S. Bartlett. J. Royal Statist. Soc. Suppl. 8 (1946), 88–90.
A. EinsteinMáthode pour la détermination de valeurs statistiques d’observations concernant des grandeurs sourmises à des fluctuations irrégulières. Arch. Sci. Phys et Natur. Ser. 37 (1914), 254–256.
V.A. EpanechnikovNonparametric estimations of a multivariate probably density. Theory of Prob. and its Applications 14 (1969), 153–158.
R. EubankSpline Smoothing and Nonparametric Regression Dekker, New York (1988).
J. FanDesign-adaptive nonparametric regression. J. Amer. Statist. Assoc. 87 (1992), 998–1004.
Th. Gasser, H.G. MüllerKernel estimation of regression functions. Smoothing techniques for curve estimation. Ed. Th. Gasser, M. Rosemblatt. Lecture notes in Mathematics 757 (1979), 23–68.
Th. Gasser, H.G. Müller, V. MammitzschKernels for nonparametric curve estimation. J. Royal Statist. Soc., B47 (1985), 238–252.
B. Granovsky, H.G. MüllerOptimizing kernel methods. A unifying variational principle. International Statistical Review 59 (1991), 373–388.
A.P. Korostelev, A.B. TsybakovMinmax Theory of Image Reconstruction. Lecture notes in Statistic 82, Springer, New York (1993).
P. Lancaster, K. SalkauskasSurfaces generated by moving least squares methods. Math. Comp. 37 (1981), 141–158.
P. Lancaster, K. SalkauskasCurve and Surface Fitting. Academic Press (1986).
M. LejeuneEstimation non-paramétrique par noyaux: régression polynomiale mobile Revue de Statistique Appliquées 33 (1985), 43–67.
K.C. LiSliced inverse regression for dimension reduction (with discussion). J. Amer. Statist. Assoc. 86 (1991), 316–327.
H.G. MüllerEmpirical bandwidth choice for nonparametric kernel regression by means of pilot estimators. Statistics and Decisions Suppl. 2 (1985), 193–206.
H.G. MüllerWeighted local regression and kernel methods for nonparametric curve fitting. J. Amer. Statist. Assoc 82 (1987), 231–238.
H.G. MüllerNonparametric regression Analysis of Longitudinal Data. Lecture Notes in Statist. 46, Springer-Verlag, New York (1988).
H.G. MüllerOn the boundary kernel method for nonparametric curve estimation near endpoints. Scandinavian J. Statist. 20 (1993), 313–328.
H.G. Müller, K. PrewittMultiparameter bandwidth processes and adaptive surface smoothing. J. Multivar. Anal. 47 (1993), 1–21.
H.G. Müller, K.S. SongMaximum estimation of multidimensional boundaries. J. Multivar. Anal. (1992), in press
H.G. MüllerCube splitting in multidimensional edge estimation. In:Change-Point Problems (E. Carlstein, H.G. Müller, D. Sigemund, eds.) IMS Lecture Notes and Monographs Series (1993), to appear.
E.A. NadarayaOn estimating regression. Theory of Prob. and its Applications 9 (1964), 141–142.
M.B. Priestley, M.T. ChaoNonparametric function fitting. J. Royal Statist. Soc., B34 (1972), 384–392.
C.H. ReinschSmoothing by spline functions. Numerical Math. 10 (1967), 177–183.
M. RosemblattCurve Estimates. Annals of Math. Statist. 42 (1971), 1815–1842.
M. RosemblattStochastic Curve Estimation. NSF-CMBS Regional Conference Series in Probability and Statistics Vol. 3 IMS, Hayward, California (1991).
M. Rudemo, H. StryhnBoundary estimation for star-shaped objects. In:Change-Point Problems (E. Carlstein, H.G. Müller, D. Sigemund, eds.) IMS Lecture Notes and Monographs Series (1993), to appear.
D. Ruppert, M. WandMultivariate locally weighted least squares regression. Preprint, Cornell University (1992).
I.J. SchoembergSpline functions and the problem of graduation. Proc. Nat’l Academy of Science 52 (1964), 947–950.
D.W. ScottMultivariate Density Estimation, Wiley, New York (1992).
H.S. ShapiroSmoothing and Approximation of Functions. Van Nostrand, New York (1969).
J. Shiau, G. Wahba, D. JohnsonPartial spline models for the inclusion of tropogause and frontal boundary information. J. Atmos. Ocean. Tech. 3 (1986), 714–725.
C.J. StoneConsistent nonparametric regression. Ann. Statist. 5 (1977), 595–620.
C.J. StoneOptimal global rates of convergence for nonparametric regression. Ann. Statist. 10 (1982), 1040–1053.
G. WahbaSpline Models for Observational Data. Society for Industrial and Applied Mathematics, Monograph (1990).
G.S. WatsonSmooth regression analysis. Sankhyā A26 (1964), 359–372.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Müller, HG. Surface and function approximation with nonparametric regression. Seminario Mat. e. Fis. di Milano 63, 171–211 (1993). https://doi.org/10.1007/BF02925100
Issue Date:
DOI: https://doi.org/10.1007/BF02925100