Multivariate spline analysis for multiplicative models: Estimation, testing and application to climate change

This paper presents multiplicative, or bi-additive, models with some spline-type regularity for a rectangular array of data, for example in space and time. The one-dimensional smoothing spline model is extended to this multiplicative model including regularity in each dimension. For estimation, we prove the existence of the maximum penalized likelihood estimates (MPLEs), and introduce a numerical algorithm that converges in a weak sense to a critical point of the penalized likelihood. Explicit MPLEs are given in two important particular cases.With regard to hypothesis testing, we focus on the "no effect" test and prove that the null distribution of the penalized likelihood ratio test (PLRT) does not depend on the nuisance parameters under H 0 , leading to easy Monte-Carlo techniques. Numerical results are presented for both simulated data and climate data. For simulated data, our estimation algorithm is shown to have a good behavior. The application to climate data illustrates how multivariate spline analysis for multiplicative models may be of interest in climate change detection.

[1]  V. Marčenko,et al.  DISTRIBUTION OF EIGENVALUES FOR SOME SETS OF RANDOM MATRICES , 1967 .

[2]  Henri Caussinus,et al.  Detection and correction of artificial shifts in climate series , 2004 .

[3]  Ren-Hong Wang,et al.  Multivariate B-splines on triangulated rectangles☆ , 1983 .

[4]  Francis W. Zwiers,et al.  Use of models in detection and attribution of climate change , 2011 .

[5]  Robert E. Schapire,et al.  The Boosting Approach to Machine Learning An Overview , 2003 .

[6]  G. Wahba,et al.  Some results on Tchebycheffian spline functions , 1971 .

[7]  Ren-Hong Wang,et al.  Multivariate spline spaces , 1983 .

[8]  Grace Wahba,et al.  Testing the (Parametric) Null Model Hypothesis in (Semiparametric) Partial and Generalized Spline Models , 1988 .

[9]  Claudia Tebaldi,et al.  Pattern scaling: Its strengths and limitations, and an update on the latest model simulations , 2014, Climatic Change.

[10]  T. D. Mitchell,et al.  Pattern Scaling: An Examination of the Accuracy of the Technique for Describing Future Climates , 2003 .

[11]  Dietrich von Rosen,et al.  The multilinear normal distribution: Introduction and some basic properties , 2013, J. Multivar. Anal..

[12]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[13]  J. Friedman.  Multivariate adaptive regression splines , 1990 .

[14]  Jan de Leeuw,et al.  Block-relaxation Algorithms in Statistics , 1994 .

[15]  C. Deser,et al.  Quantifying the Role of Internal Climate Variability in Future Climate Trends , 2015 .

[16]  Jean-Baptiste Denis,et al.  Asymptotic Confidence Regions for Biadditive Models: Interpreting Genotype‐Environment Interactions , 1996 .

[17]  Bernhard Schölkopf,et al.  A Generalized Representer Theorem , 2001, COLT/EuroCOLT.

[18]  Corinne Le Quéré,et al.  Climate Change 2013: The Physical Science Basis , 2013 .

[19]  M. Wand,et al.  Exact likelihood ratio tests for penalised splines , 2005 .

[20]  D. Zimmerman,et al.  The likelihood ratio test for a separable covariance matrix , 2005 .

[21]  S. Planton,et al.  A method for regional climate change detection using smooth temporal patterns , 2010 .

[22]  Yuedong Wang,et al.  Hypothesis testing in smoothing spline models , 2004 .

[23]  Camille Parmesan,et al.  Good practice guidance paper on detection and attribution related to anthropogenic climate change , 2010 .

[24]  G. Wahba.  Spline models for observational data , 1990 .