An assessment of finite sample performance of adaptive methods in density estimation

Abstract Kernel density estimation is a powerful tool for exploratory data analysis. Adaptive methods can improve the appearance of these curve estimates by smoothing away spurious “wiggles”. The finite sample performance of several location dependent bandwidths is studied by simulation. The mean integrated squared error (MISE) of the adaptive methods is compared to the MISE of a well-respected constant bandwidth often referred to as the Sheather–Jones plug-in (SJPI). A surprising fact is that the MISE performance of the SJPI is often quite close to that of the adaptive methods. In addition, an alternative visual error criterion is used to rate performance as an experienced data analyst might. Many interesting questions concerning the implementation of these adaptive approaches are addressed.

[1]  J. Marron,et al.  Improved Variable Window Kernel Estimates of Probability Densities , 1995 .

[2]  Ian Mckay.  A note on bias reduction in variable‐kernel density estimates , 1993 .

[3]  B. Silverman.  Density estimation for statistics and data analysis , 1986 .

[4]  W. R. Schucany,et al.  Adaptive Bandwidth Choice for Kernel Regression , 1995 .

[5]  A. Cuevas,et al.  A comparative study of several smoothing methods in density estimation , 1994 .

[6]  Jianqing Fan,et al.  Fast Implementations of Nonparametric Curve Estimators , 1994 .

[7]  Joachim Engel,et al.  An iterative bandwidth selector for kernel estimation of densities and their derivatives , 1994 .

[8]  J. Staniswalis.  Local Bandwidth Selection for Kernel Estimates , 1989 .

[9]  Jan Mielniczuk,et al.  Local data-driven bandwidth choice for density estimation , 1989 .

[10]  Bandwidth selection procedures tor kernel density estimates , 1982 .

[11]  S. Sheather.  A data-based algorithm for choosing the window width when estimating the density at a point , 1983 .

[12]  D. W. Scott,et al.  On Locally Adaptive Density Estimation , 1996 .

[13]  A. Q. D. Río,et al.  A local cross-validation algorithm for dependent data , 1992 .

[14]  James Stephen Marron,et al.  Visual Error Criteria for Qualitative Smoothing , 1995 .

[15]  T. Gasser,et al.  Locally Adaptive Bandwidth Choice for Kernel Regression Estimators , 1993 .

[16]  W. Cleveland,et al.  Computational methods for local regression , 1991 .

[17]  G. Terrell.  The Maximal Smoothing Principle in Density Estimation , 1990 .

[18]  L. Breiman,et al.  Variable Kernel Estimates of Multivariate Densities , 1977 .

[19]  Martin L. Hazelton.  Bandwidth Selection for Local Density Estimators , 1996 .

[20]  J. Marron,et al.  Progress in data-based bandwidth selection for kernel density estimation , 1996 .

[21]  B. Turlach,et al.  Rejoinder to ``Practical performance of several data driven bandwidth selectors" , 1992 .

[22]  M. Woodroofe.  On Choosing a Delta-Sequence , 1970 .

[23]  James Stephen Marron,et al.  On the Amount of Noise Inherent in Bandwidth Selection for a Kernel Density Estimator , 1987 .

[24]  A. Bowman,et al.  Adaptive Smoothing and Density-Based Tests of Multivariate Normality , 1993 .

[25]  James Stephen Marron,et al.  Estimation of integrated squared density derivatives , 1987 .

[26]  Peter Hall,et al.  A local cross-validation algorithm , 1989 .

[27]  James Stephen Marron,et al.  A simple root n bandwidth selector , 1991 .

[28]  D. W. Scott,et al.  Oversmoothed Nonparametric Density Estimates , 1985 .

[29]  D. W. Scott,et al.  Biased and Unbiased Cross-Validation in Density Estimation , 1987 .

[30]  H. Müller,et al.  Variable Bandwidth Kernel Estimators of Regression Curves , 1987 .

[31]  J. Friedman.  A VARIABLE SPAN SMOOTHER , 1984 .

[32]  Peter Hall,et al.  On local smoothing of nonparametric curve estimators , 1996 .

[33]  Simon J. Sheather,et al.  Using non stochastic terms to advantage in kernel-based estimation of integrated squared density derivatives , 1991 .

[34]  A. Krieger,et al.  Weak Convergence and Efficient Density Estimation at a Point , 1981 .

[35]  P. Hall.  On Kullback-Leibler loss and density estimation , 1987 .

[36]  W. R. Schucany.  Locally optimal window widths for kernel density estimation with large samples , 1989 .

[37]  M. C. Jones.  Variable kernel density estimates and variable kernel density estimates , 1990 .

[38]  James Stephen Marron,et al.  Lower bounds for bandwidth selection in density estimation , 1991 .

[39]  Ian Abramson.  On Bandwidth Variation in Kernel Estimates-A Square Root Law , 1982 .

[40]  William S. Cleveland,et al.  Visualizing Data , 1993 .

[41]  D. W. Scott,et al.  Variable Kernel Density Estimation , 1992 .

[42]  A. Bowman.  A comparative study of some kernel-based nonparametric density estimators , 1985 .

[43]  J. Marron.  An Asymptotically Efficient Solution to the Bandwidth Problem of Kernel Density Estimation , 1985 .

[44]  Luc Devroye,et al.  On the Relationship Between Stability of Extreme Order Statistics and Convergence of the Maximum Likelihood Kernel Density Estimate , 1989 .

[45]  James Stephen Marron,et al.  Canonical kernels for density estimation , 1988 .

[46]  H. Müller,et al.  Kernels for Nonparametric Curve Estimation , 1985 .

[47]  M. C. Jones,et al.  A reliable data-based bandwidth selection method for kernel density estimation , 1991 .

[48]  M. C. Jones,et al.  Universal smoothing factor selection in density estimation: theory and practice , 1997 .