New Developments in Sparse PLS Regression
Methods based on partial least squares (PLS) regression, which has recently gained much attention in the analysis of high-dimensional genomic datasets, have been developed since the early 2000s for performing variable selection. Most of these techniques rely on tuning parameters that are often deter...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Frontiers Media S.A.
2021-07-01
|
Series: | Frontiers in Applied Mathematics and Statistics |
Subjects: | |
Online Access: | https://www.frontiersin.org/articles/10.3389/fams.2021.693126/full |
id |
doaj-0a8ab28962a346b5bc74e0bb4189d93c |
---|---|
record_format |
Article |
spelling |
doaj-0a8ab28962a346b5bc74e0bb4189d93c2021-07-16T11:27:55ZengFrontiers Media S.A.Frontiers in Applied Mathematics and Statistics2297-46872021-07-01710.3389/fams.2021.693126693126New Developments in Sparse PLS RegressionJérémy Magnanensi0Jérémy Magnanensi1Myriam Maumy-Bertrand2Myriam Maumy-Bertrand3Nicolas Meyer4Frédéric Bertrand5Frédéric Bertrand6IRMA, CNRS UMR 7501, Labex IRMIA, Université de Strasbourg and CNRS, Strasbourg, FranceLaboratoire de Biostatistique et Informatique Médicale, Faculté de Médecine, EA3430, Université de Strasbourg and CNRS, Strasbourg, FranceIRMA, CNRS UMR 7501, Labex IRMIA, Université de Strasbourg and CNRS, Strasbourg, FranceLIST3N, Université de Technologie de Troyes, Troyes, FranceLaboratoire de Biostatistique et Informatique Médicale, Faculté de Médecine, EA3430, Université de Strasbourg and CNRS, Strasbourg, FranceIRMA, CNRS UMR 7501, Labex IRMIA, Université de Strasbourg and CNRS, Strasbourg, FranceLIST3N, Université de Technologie de Troyes, Troyes, FranceMethods based on partial least squares (PLS) regression, which has recently gained much attention in the analysis of high-dimensional genomic datasets, have been developed since the early 2000s for performing variable selection. Most of these techniques rely on tuning parameters that are often determined by cross-validation (CV) based methods, which raises essential stability issues. To overcome this, we have developed a new dynamic bootstrap-based method for significant predictor selection, suitable for both PLS regression and its incorporation into generalized linear models (GPLS). It relies on establishing bootstrap confidence intervals, which allows testing of the significance of predictors at preset type I risk α, and avoids CV. We have also developed adapted versions of sparse PLS (SPLS) and sparse GPLS regression (SGPLS), using a recently introduced non-parametric bootstrap-based technique to determine the numbers of components. We compare their variable selection reliability and stability concerning tuning parameters determination and their predictive ability, using simulated data for PLS and real microarray gene expression data for PLS-logistic classification. We observe that our new dynamic bootstrap-based method has the property of best separating random noise in y from the relevant information with respect to other methods, leading to better accuracy and predictive abilities, especially for non-negligible noise levels.https://www.frontiersin.org/articles/10.3389/fams.2021.693126/fullvariable selectionpartial least squaressparse partial least squaresgeneralized partial least squaresbootstrapstability |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Jérémy Magnanensi Jérémy Magnanensi Myriam Maumy-Bertrand Myriam Maumy-Bertrand Nicolas Meyer Frédéric Bertrand Frédéric Bertrand |
spellingShingle |
Jérémy Magnanensi Jérémy Magnanensi Myriam Maumy-Bertrand Myriam Maumy-Bertrand Nicolas Meyer Frédéric Bertrand Frédéric Bertrand New Developments in Sparse PLS Regression Frontiers in Applied Mathematics and Statistics variable selection partial least squares sparse partial least squares generalized partial least squares bootstrap stability |
author_facet |
Jérémy Magnanensi Jérémy Magnanensi Myriam Maumy-Bertrand Myriam Maumy-Bertrand Nicolas Meyer Frédéric Bertrand Frédéric Bertrand |
author_sort |
Jérémy Magnanensi |
title |
New Developments in Sparse PLS Regression |
title_short |
New Developments in Sparse PLS Regression |
title_full |
New Developments in Sparse PLS Regression |
title_fullStr |
New Developments in Sparse PLS Regression |
title_full_unstemmed |
New Developments in Sparse PLS Regression |
title_sort |
new developments in sparse pls regression |
publisher |
Frontiers Media S.A. |
series |
Frontiers in Applied Mathematics and Statistics |
issn |
2297-4687 |
publishDate |
2021-07-01 |
description |
Methods based on partial least squares (PLS) regression, which has recently gained much attention in the analysis of high-dimensional genomic datasets, have been developed since the early 2000s for performing variable selection. Most of these techniques rely on tuning parameters that are often determined by cross-validation (CV) based methods, which raises essential stability issues. To overcome this, we have developed a new dynamic bootstrap-based method for significant predictor selection, suitable for both PLS regression and its incorporation into generalized linear models (GPLS). It relies on establishing bootstrap confidence intervals, which allows testing of the significance of predictors at preset type I risk α, and avoids CV. We have also developed adapted versions of sparse PLS (SPLS) and sparse GPLS regression (SGPLS), using a recently introduced non-parametric bootstrap-based technique to determine the numbers of components. We compare their variable selection reliability and stability concerning tuning parameters determination and their predictive ability, using simulated data for PLS and real microarray gene expression data for PLS-logistic classification. We observe that our new dynamic bootstrap-based method has the property of best separating random noise in y from the relevant information with respect to other methods, leading to better accuracy and predictive abilities, especially for non-negligible noise levels. |
topic |
variable selection partial least squares sparse partial least squares generalized partial least squares bootstrap stability |
url |
https://www.frontiersin.org/articles/10.3389/fams.2021.693126/full |
work_keys_str_mv |
AT jeremymagnanensi newdevelopmentsinsparseplsregression AT jeremymagnanensi newdevelopmentsinsparseplsregression AT myriammaumybertrand newdevelopmentsinsparseplsregression AT myriammaumybertrand newdevelopmentsinsparseplsregression AT nicolasmeyer newdevelopmentsinsparseplsregression AT fredericbertrand newdevelopmentsinsparseplsregression AT fredericbertrand newdevelopmentsinsparseplsregression |
_version_ |
1721297608038678528 |