On the Choice of Difference Sequence in a Unified Framework for Variance Estimation in Nonparametric Regression

Handle URI:
http://hdl.handle.net/10754/625520
Title:
On the Choice of Difference Sequence in a Unified Framework for Variance Estimation in Nonparametric Regression
Authors:
Dai, Wenlin; Tong, Tiejun; Zhu, Lixing
Abstract:
Difference-based methods do not require estimating the mean function in nonparametric regression and are therefore popular in practice. In this paper, we propose a unified framework for variance estimation that combines the linear regression method with the higher-order difference estimators systematically. The unified framework has greatly enriched the existing literature on variance estimation that includes most existing estimators as special cases. More importantly, the unified framework has also provided a smart way to solve the challenging difference sequence selection problem that remains a long-standing controversial issue in nonparametric regression for several decades. Using both theory and simulations, we recommend to use the ordinary difference sequence in the unified framework, no matter if the sample size is small or if the signal-to-noise ratio is large. Finally, to cater for the demands of the application, we have developed a unified R package, named VarED, that integrates the existing difference-based estimators and the unified estimators in nonparametric regression and have made it freely available in the R statistical program http://cran.r-project.org/web/packages/.
KAUST Department:
Computer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
Citation:
Dai W, Tong T, Zhu L (2017) On the Choice of Difference Sequence in a Unified Framework for Variance Estimation in Nonparametric Regression. Statistical Science 32: 455–468. Available: http://dx.doi.org/10.1214/17-STS613.
Publisher:
Institute of Mathematical Statistics
Journal:
Statistical Science
Issue Date:
1-Sep-2017
DOI:
10.1214/17-STS613
Type:
Article
ISSN:
0883-4237
Sponsors:
Tiejun Tong's research was supported by the National Natural Science Foundation of China grant (No. 11671338), and the Hong Kong Baptist University grants FRG1/14-15/044, FRG2/15-16/019 and FRG2/15-16/038. Lixing Zhu's research was supported by the Hong Kong Research Grants Council grant (No. HKBU202810). The authors thank the editor, the associate editor and two reviewers for their constructive comments that have led to a substantial improvement of the paper.
Additional Links:
https://projecteuclid.org/euclid.ss/1504253126
Appears in Collections:
Articles; Computer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division

Full metadata record

DC FieldValue Language
dc.contributor.authorDai, Wenlinen
dc.contributor.authorTong, Tiejunen
dc.contributor.authorZhu, Lixingen
dc.date.accessioned2017-10-01T05:07:27Z-
dc.date.available2017-10-01T05:07:27Z-
dc.date.issued2017-09-01en
dc.identifier.citationDai W, Tong T, Zhu L (2017) On the Choice of Difference Sequence in a Unified Framework for Variance Estimation in Nonparametric Regression. Statistical Science 32: 455–468. Available: http://dx.doi.org/10.1214/17-STS613.en
dc.identifier.issn0883-4237en
dc.identifier.doi10.1214/17-STS613en
dc.identifier.urihttp://hdl.handle.net/10754/625520-
dc.description.abstractDifference-based methods do not require estimating the mean function in nonparametric regression and are therefore popular in practice. In this paper, we propose a unified framework for variance estimation that combines the linear regression method with the higher-order difference estimators systematically. The unified framework has greatly enriched the existing literature on variance estimation that includes most existing estimators as special cases. More importantly, the unified framework has also provided a smart way to solve the challenging difference sequence selection problem that remains a long-standing controversial issue in nonparametric regression for several decades. Using both theory and simulations, we recommend to use the ordinary difference sequence in the unified framework, no matter if the sample size is small or if the signal-to-noise ratio is large. Finally, to cater for the demands of the application, we have developed a unified R package, named VarED, that integrates the existing difference-based estimators and the unified estimators in nonparametric regression and have made it freely available in the R statistical program http://cran.r-project.org/web/packages/.en
dc.description.sponsorshipTiejun Tong's research was supported by the National Natural Science Foundation of China grant (No. 11671338), and the Hong Kong Baptist University grants FRG1/14-15/044, FRG2/15-16/019 and FRG2/15-16/038. Lixing Zhu's research was supported by the Hong Kong Research Grants Council grant (No. HKBU202810). The authors thank the editor, the associate editor and two reviewers for their constructive comments that have led to a substantial improvement of the paper.en
dc.publisherInstitute of Mathematical Statisticsen
dc.relation.urlhttps://projecteuclid.org/euclid.ss/1504253126en
dc.rightsArchived with thanks to Statistical Scienceen
dc.subjectDifference-based estimatoren
dc.subjectnonparametric regressionen
dc.subjectoptimal difference sequenceen
dc.subjectordinary difference sequenceen
dc.subjectresidual varianceen
dc.titleOn the Choice of Difference Sequence in a Unified Framework for Variance Estimation in Nonparametric Regressionen
dc.typeArticleen
dc.contributor.departmentComputer, Electrical and Mathematical Sciences and Engineering (CEMSE) Divisionen
dc.identifier.journalStatistical Scienceen
dc.eprint.versionPublisher's Version/PDFen
dc.contributor.institutionDepartment of Mathematics, Hong Kong Baptist University, , Hong Kongen
kaust.authorDai, Wenlinen
All Items in KAUST are protected by copyright, with all rights reserved, unless otherwise indicated.