A frisch-newton algorithm for sparse quantile regression

Roger Koenker, Pin T Ng

Research output: Contribution to journalArticle

21 Citations (Scopus)

Abstract

Recent experience has shown that interior-point methods using a log barrier approach are far superior to classical simplex methods for computing solutions to large parametric quantile regression problems. In many large empirical applications, the design matrix has a very sparse structure. A typical example is the classical fixed-effect model for panel data where the parametric dimension of the model can be quite large, but the number of non-zero elements is quite small. Adopting recent developments in sparse linear algebra we introduce a modified version of the Frisch-Newton algorithm for quantile regression described in Portnoy and Koenker[28]. The new algorithm substantially reduces the storage (memory) requirements and increases computational speed. The modified algorithm also facilitates the development of nonparametric quantile regression methods. The pseudo design matrices employed in nonparametric quantile regression smoothing are inherently sparse in both the fidelity and roughness penalty components. Exploiting the sparse structure of these problems opens up a whole range of new possibilities for multivariate smoothing on large data sets via ANOVA-type decomposition and partial linear models.

Original languageEnglish (US)
Pages (from-to)225-236
Number of pages12
JournalActa Mathematicae Applicatae Sinica
Volume21
Issue number2
DOIs
StatePublished - May 2005

Fingerprint

Quantile Regression
Nonparametric Regression
Smoothing
Linear algebra
Roughness Penalty
Partial Linear Model
Analysis of variance (ANOVA)
Fixed Effects Model
Parametric Regression
Simplex Method
Interior Point Method
Panel Data
Surface roughness
Large Data Sets
Fidelity
Decomposition
Data storage equipment
Open Problems
Decompose
Computing

Keywords

  • Interior-point algorithm
  • Quantile regression
  • Sparse linear algebra

ASJC Scopus subject areas

  • Applied Mathematics

Cite this

A frisch-newton algorithm for sparse quantile regression. / Koenker, Roger; Ng, Pin T.

In: Acta Mathematicae Applicatae Sinica, Vol. 21, No. 2, 05.2005, p. 225-236.

Research output: Contribution to journalArticle

@article{875b525b13f644a6b7b7f8d8358f6883,
title = "A frisch-newton algorithm for sparse quantile regression",
abstract = "Recent experience has shown that interior-point methods using a log barrier approach are far superior to classical simplex methods for computing solutions to large parametric quantile regression problems. In many large empirical applications, the design matrix has a very sparse structure. A typical example is the classical fixed-effect model for panel data where the parametric dimension of the model can be quite large, but the number of non-zero elements is quite small. Adopting recent developments in sparse linear algebra we introduce a modified version of the Frisch-Newton algorithm for quantile regression described in Portnoy and Koenker[28]. The new algorithm substantially reduces the storage (memory) requirements and increases computational speed. The modified algorithm also facilitates the development of nonparametric quantile regression methods. The pseudo design matrices employed in nonparametric quantile regression smoothing are inherently sparse in both the fidelity and roughness penalty components. Exploiting the sparse structure of these problems opens up a whole range of new possibilities for multivariate smoothing on large data sets via ANOVA-type decomposition and partial linear models.",
keywords = "Interior-point algorithm, Quantile regression, Sparse linear algebra",
author = "Roger Koenker and Ng, {Pin T}",
year = "2005",
month = "5",
doi = "10.1007/s10255-005-0231-1",
language = "English (US)",
volume = "21",
pages = "225--236",
journal = "Acta Mathematicae Applicatae Sinica",
issn = "0168-9673",
publisher = "Springer Verlag",
number = "2",

}

TY - JOUR

T1 - A frisch-newton algorithm for sparse quantile regression

AU - Koenker, Roger

AU - Ng, Pin T

PY - 2005/5

Y1 - 2005/5

N2 - Recent experience has shown that interior-point methods using a log barrier approach are far superior to classical simplex methods for computing solutions to large parametric quantile regression problems. In many large empirical applications, the design matrix has a very sparse structure. A typical example is the classical fixed-effect model for panel data where the parametric dimension of the model can be quite large, but the number of non-zero elements is quite small. Adopting recent developments in sparse linear algebra we introduce a modified version of the Frisch-Newton algorithm for quantile regression described in Portnoy and Koenker[28]. The new algorithm substantially reduces the storage (memory) requirements and increases computational speed. The modified algorithm also facilitates the development of nonparametric quantile regression methods. The pseudo design matrices employed in nonparametric quantile regression smoothing are inherently sparse in both the fidelity and roughness penalty components. Exploiting the sparse structure of these problems opens up a whole range of new possibilities for multivariate smoothing on large data sets via ANOVA-type decomposition and partial linear models.

AB - Recent experience has shown that interior-point methods using a log barrier approach are far superior to classical simplex methods for computing solutions to large parametric quantile regression problems. In many large empirical applications, the design matrix has a very sparse structure. A typical example is the classical fixed-effect model for panel data where the parametric dimension of the model can be quite large, but the number of non-zero elements is quite small. Adopting recent developments in sparse linear algebra we introduce a modified version of the Frisch-Newton algorithm for quantile regression described in Portnoy and Koenker[28]. The new algorithm substantially reduces the storage (memory) requirements and increases computational speed. The modified algorithm also facilitates the development of nonparametric quantile regression methods. The pseudo design matrices employed in nonparametric quantile regression smoothing are inherently sparse in both the fidelity and roughness penalty components. Exploiting the sparse structure of these problems opens up a whole range of new possibilities for multivariate smoothing on large data sets via ANOVA-type decomposition and partial linear models.

KW - Interior-point algorithm

KW - Quantile regression

KW - Sparse linear algebra

UR - http://www.scopus.com/inward/record.url?scp=18244365578&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=18244365578&partnerID=8YFLogxK

U2 - 10.1007/s10255-005-0231-1

DO - 10.1007/s10255-005-0231-1

M3 - Article

AN - SCOPUS:18244365578

VL - 21

SP - 225

EP - 236

JO - Acta Mathematicae Applicatae Sinica

JF - Acta Mathematicae Applicatae Sinica

SN - 0168-9673

IS - 2

ER -