Revisiting fitting monotone polynomials to data

We revisit Hawkins’ (Comput Stat 9(3):233–247, 1994) algorithm for fitting monotonic polynomials and discuss some practical issues that we encountered using this algorithm, for example when fitting high degree polynomials or situations with a sparse design matrix but multiple observations per $$x$$-value. As an alternative, we describe a new approach to fitting monotone polynomials to data, based on different characterisations of monotone polynomials and using a Levenberg–Marquardt type algorithm. We consider different parameterisations, examine effective starting values for the non-linear algorithms, and discuss some limitations. We illustrate our methodology with examples of simulated and real world data. All algorithms discussed in this paper are available in the R Development Core Team (A language and environment for statistical computing, R Foundation for Statistical Computing, Vienna, 2011) package MonoPoly.

[1]  James O. Ramsay,et al.  Applied Functional Data Analysis: Methods and Case Studies , 2002 .

[2]  K. Rösler,et al.  The latency distribution of motor evoked potentials in patients with multiple sclerosis , 2012, Clinical Neurophysiology.

[3]  J. Ramsay.  Estimating smooth monotone functions , 1998 .

[4]  M. R. Osborne.  Nonlinear least squares — the Levenberg algorithm revisited , 1976, The Journal of the Australian Mathematical Society. Series B. Applied Mathematics.

[5]  Berwin A. Turlach,et al.  Shape constrained smoothing using smoothing splines , 2005, Comput. Stat..

[6]  D. Goldfarb,et al.  Dual and primal-dual methods for solving strictly convex quadratic programs , 1982 .

[7]  M. Powell.  The BOBYQA algorithm for bound constrained optimization without derivatives , 2009 .

[8]  Holger Dette,et al.  A comparative study of monotone nonparametric kernel estimates , 2006 .

[9]  John A. Nelder,et al.  A Simplex Method for Function Minimization , 1965, Comput. J..

[10]  Sujit K. Ghosh,et al.  A variable selection approach to monotonic regression with Bernstein polynomials , 2011 .

[11]  Donald Goldfarb,et al.  A numerically stable dual method for solving strictly convex quadratic programs , 1983, Math. Program..

[12]  Trevor Hastie,et al.  Regularization Paths for Generalized Linear Models via Coordinate Descent. , 2010, Journal of statistical software.

[13]  R. Tibshirani,et al.  The Monotone Smoothing of Scatterplots , 1984 .

[14]  Martin L. Hazelton,et al.  Semiparametric regression with shape-constrained penalized splines , 2011, Comput. Stat. Data Anal..

[15]  L. Fausett.  Numerical Methods: Algorithms and Applications , 2002 .

[16]  P. Hall,et al.  NONPARAMETRIC KERNEL REGRESSION SUBJECT TO MONOTONICITY CONSTRAINTS , 2001 .

[17]  Douglas M. Hawkins,et al.  Fitting Monotonic Polynomials to Data , 1994 .

[18]  H. Dette,et al.  A simple nonparametric estimator of a strictly monotone regression function , 2006 .

[19]  R. Tibshirani,et al.  PATHWISE COORDINATE OPTIMIZATION , 2007, 0708.1485.

[20]  C. Elphinstone,et al.  A target distribution model for nonparametric density estimation , 1983 .

[21]  Enno Mammen,et al.  Estimating a Smooth Monotone Regression Function , 1991 .

[22]  Dominik Heinzmann,et al.  A filtered polynomial approach to density estimation , 2008, Comput. Stat..

[23]  K. Rösler,et al.  A method to measure the distribution of latencies of motor evoked potentials in man , 2011, Clinical Neurophysiology.

[24]  E. Mammen,et al.  A General Projection Framework for Constrained Smoothing , 2001 .