Collaborative sliced inverse regression

Alessandro Chiancone, Stéphane Girard, Jocelyn Chanussot

Research output: Contribution to journalArticle

Abstract

Sliced inverse regression (SIR) is an effective method for dimensionality reduction in high-dimensional regression problems. However, the method has requirements on the distribution of the predictors that are hard to check since they depend on unobserved variables. It has been shown that, if the distribution of the predictors is elliptical, then these requirements are satisfied. In case of mixture models, the ellipticity is violated and in addition there is no assurance of a single underlying regression model among the different components. Our approach clusterizes the predictors space to force the condition to hold on each cluster and includes a merging technique to look for different underlying models in the data. A study on simulated data as well as two real applications are provided. It appears that SIR, unsurprisingly, is not capable of dealing with a mixture of Gaussians involving different underlying models whereas our approach is able to correctly investigate the mixture.

LanguageEnglish
Pages6035-6053
Number of pages19
JournalCommunications in Statistics - Theory and Methods
Volume46
Issue number12
DOIs
StatusPublished - 18 Jun 2017

Fingerprint

Sliced Inverse Regression
Predictors
Ellipticity
Requirements
Dimensionality Reduction
Mixture Model
Merging
Regression Model
High-dimensional
Regression
Model

Keywords

  • Inverse regression
  • mixture models
  • sufficient dimension regression

ASJC Scopus subject areas

  • Statistics and Probability

Cite this

Collaborative sliced inverse regression. / Chiancone, Alessandro; Girard, Stéphane; Chanussot, Jocelyn.

In: Communications in Statistics - Theory and Methods, Vol. 46, No. 12, 18.06.2017, p. 6035-6053.

Research output: Contribution to journalArticle

Chiancone, Alessandro ; Girard, Stéphane ; Chanussot, Jocelyn. / Collaborative sliced inverse regression. In: Communications in Statistics - Theory and Methods. 2017 ; Vol. 46, No. 12. pp. 6035-6053
@article{da73f5262ba642948fa10440a9d0b427,
title = "Collaborative sliced inverse regression",
abstract = "Sliced inverse regression (SIR) is an effective method for dimensionality reduction in high-dimensional regression problems. However, the method has requirements on the distribution of the predictors that are hard to check since they depend on unobserved variables. It has been shown that, if the distribution of the predictors is elliptical, then these requirements are satisfied. In case of mixture models, the ellipticity is violated and in addition there is no assurance of a single underlying regression model among the different components. Our approach clusterizes the predictors space to force the condition to hold on each cluster and includes a merging technique to look for different underlying models in the data. A study on simulated data as well as two real applications are provided. It appears that SIR, unsurprisingly, is not capable of dealing with a mixture of Gaussians involving different underlying models whereas our approach is able to correctly investigate the mixture.",
keywords = "Inverse regression, mixture models, sufficient dimension regression",
author = "Alessandro Chiancone and St{\'e}phane Girard and Jocelyn Chanussot",
year = "2017",
month = "6",
day = "18",
doi = "10.1080/03610926.2015.1116578",
language = "English",
volume = "46",
pages = "6035--6053",
journal = "Communications in statistics / Theory and methods",
issn = "0361-0926",
publisher = "Taylor and Francis Ltd.",
number = "12",

}

TY - JOUR

T1 - Collaborative sliced inverse regression

AU - Chiancone,Alessandro

AU - Girard,Stéphane

AU - Chanussot,Jocelyn

PY - 2017/6/18

Y1 - 2017/6/18

N2 - Sliced inverse regression (SIR) is an effective method for dimensionality reduction in high-dimensional regression problems. However, the method has requirements on the distribution of the predictors that are hard to check since they depend on unobserved variables. It has been shown that, if the distribution of the predictors is elliptical, then these requirements are satisfied. In case of mixture models, the ellipticity is violated and in addition there is no assurance of a single underlying regression model among the different components. Our approach clusterizes the predictors space to force the condition to hold on each cluster and includes a merging technique to look for different underlying models in the data. A study on simulated data as well as two real applications are provided. It appears that SIR, unsurprisingly, is not capable of dealing with a mixture of Gaussians involving different underlying models whereas our approach is able to correctly investigate the mixture.

AB - Sliced inverse regression (SIR) is an effective method for dimensionality reduction in high-dimensional regression problems. However, the method has requirements on the distribution of the predictors that are hard to check since they depend on unobserved variables. It has been shown that, if the distribution of the predictors is elliptical, then these requirements are satisfied. In case of mixture models, the ellipticity is violated and in addition there is no assurance of a single underlying regression model among the different components. Our approach clusterizes the predictors space to force the condition to hold on each cluster and includes a merging technique to look for different underlying models in the data. A study on simulated data as well as two real applications are provided. It appears that SIR, unsurprisingly, is not capable of dealing with a mixture of Gaussians involving different underlying models whereas our approach is able to correctly investigate the mixture.

KW - Inverse regression

KW - mixture models

KW - sufficient dimension regression

UR - http://www.scopus.com/inward/record.url?scp=85014750658&partnerID=8YFLogxK

U2 - 10.1080/03610926.2015.1116578

DO - 10.1080/03610926.2015.1116578

M3 - Article

VL - 46

SP - 6035

EP - 6053

JO - Communications in statistics / Theory and methods

T2 - Communications in statistics / Theory and methods

JF - Communications in statistics / Theory and methods

SN - 0361-0926

IS - 12

ER -