MARS: Selecting basis functions and knots with an empirical Bayes method

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

An empirical Bayes method to select basis functions and knots in multivariate adaptive regression spline (MARS) is proposed, which takes both advantages of frequentist model selection approaches and Bayesian approaches. A penalized likelihood is maximized to estimate regression coefficients for selected basis functions, and an approximated marginal likelihood is maximized to select knots and variables involved in basis functions. Moreover, the Akaike Bayes information criterion (ABIC) is used to determine the number of basis functions. It is shown that the proposed method gives estimation of regression structure that is relatively parsimonious and more stable for some example data sets.

Original languageEnglish
Pages (from-to)583-597
Number of pages15
JournalComputational Statistics
Volume22
Issue number4
DOIs
Publication statusPublished - Dec 2007
Externally publishedYes

Fingerprint

Multivariate Adaptive Regression Splines
Empirical Bayes Method
Spline Functions
Splines
Knot
Basis Functions
Bayes Information Criterion
Marginal Likelihood
Penalized Likelihood
Akaike Information Criterion
Regression Coefficient
Bayesian Approach
Model Selection
Regression
Empirical Bayes
Spline regression
Estimate

Keywords

  • Akaike Bayes information criterion
  • Estimation of interaction terms
  • Marginal likelihood
  • Multivariate adaptive regression
  • Penalized likelihood approach

ASJC Scopus subject areas

  • Statistics and Probability
  • Computational Mathematics

Cite this

MARS : Selecting basis functions and knots with an empirical Bayes method. / Sakamoto, Wataru.

In: Computational Statistics, Vol. 22, No. 4, 12.2007, p. 583-597.

Research output: Contribution to journalArticle

@article{4ff4f5b486f24464a9079c301d75e001,
title = "MARS: Selecting basis functions and knots with an empirical Bayes method",
abstract = "An empirical Bayes method to select basis functions and knots in multivariate adaptive regression spline (MARS) is proposed, which takes both advantages of frequentist model selection approaches and Bayesian approaches. A penalized likelihood is maximized to estimate regression coefficients for selected basis functions, and an approximated marginal likelihood is maximized to select knots and variables involved in basis functions. Moreover, the Akaike Bayes information criterion (ABIC) is used to determine the number of basis functions. It is shown that the proposed method gives estimation of regression structure that is relatively parsimonious and more stable for some example data sets.",
keywords = "Akaike Bayes information criterion, Estimation of interaction terms, Marginal likelihood, Multivariate adaptive regression, Penalized likelihood approach",
author = "Wataru Sakamoto",
year = "2007",
month = "12",
doi = "10.1007/s00180-007-0075-7",
language = "English",
volume = "22",
pages = "583--597",
journal = "Computational Statistics",
issn = "0943-4062",
publisher = "Springer Verlag",
number = "4",

}

TY - JOUR

T1 - MARS

T2 - Selecting basis functions and knots with an empirical Bayes method

AU - Sakamoto, Wataru

PY - 2007/12

Y1 - 2007/12

N2 - An empirical Bayes method to select basis functions and knots in multivariate adaptive regression spline (MARS) is proposed, which takes both advantages of frequentist model selection approaches and Bayesian approaches. A penalized likelihood is maximized to estimate regression coefficients for selected basis functions, and an approximated marginal likelihood is maximized to select knots and variables involved in basis functions. Moreover, the Akaike Bayes information criterion (ABIC) is used to determine the number of basis functions. It is shown that the proposed method gives estimation of regression structure that is relatively parsimonious and more stable for some example data sets.

AB - An empirical Bayes method to select basis functions and knots in multivariate adaptive regression spline (MARS) is proposed, which takes both advantages of frequentist model selection approaches and Bayesian approaches. A penalized likelihood is maximized to estimate regression coefficients for selected basis functions, and an approximated marginal likelihood is maximized to select knots and variables involved in basis functions. Moreover, the Akaike Bayes information criterion (ABIC) is used to determine the number of basis functions. It is shown that the proposed method gives estimation of regression structure that is relatively parsimonious and more stable for some example data sets.

KW - Akaike Bayes information criterion

KW - Estimation of interaction terms

KW - Marginal likelihood

KW - Multivariate adaptive regression

KW - Penalized likelihood approach

UR - http://www.scopus.com/inward/record.url?scp=36849047234&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=36849047234&partnerID=8YFLogxK

U2 - 10.1007/s00180-007-0075-7

DO - 10.1007/s00180-007-0075-7

M3 - Article

AN - SCOPUS:36849047234

VL - 22

SP - 583

EP - 597

JO - Computational Statistics

JF - Computational Statistics

SN - 0943-4062

IS - 4

ER -