Bayesian group selection with non-local priors

Weibing Li; Thierry Chekouo

doi:10.1007/s00180-021-01115-1

Bayesian group selection with non-local priors

Weibing Li, Thierry Chekouo

Research output: Contribution to journal › Article › peer-review

2 Scopus citations

Abstract

In many applications, variables or features can be naturally partitioned into different groups. In this article, we propose a new Bayesian hierarchical model for group selection problem when the group structure is known. We use spike and slab priors for regression coefficients, and the slab component is assumed to come from the family of nonlocal priors. Contrary to local priors commonly used in Bayesian group selection, nonlocal density priors vanish when a regression coefficient in the model is zero. We use simulation studies to assess the performance of our method and apply it to data collected from individuals undergoing cardiac catheterization at Duke University Medical center between 2001 and 2010.

Original language	English (US)
Pages (from-to)	287-302
Number of pages	16
Journal	Computational Statistics
Volume	37
Issue number	1
DOIs	https://doi.org/10.1007/s00180-021-01115-1
State	Published - Mar 2022
Externally published	Yes

Bibliographical note

Funding Information:
Thierry Chekouo is partially supported by NSERC Discovery Grants Number RGPIN-2019-04810. The content is solely the responsibility of the authors and does not necessarily represent the official views of NSERC.

Publisher Copyright:
© 2021, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.

Keywords

Bayesian computing
CATHGEN
MCMC
Spike and slab priors

Access

10.1007/s00180-021-01115-1

OpenUrl availability

Full text

Cite this

@article{b3d8d7c350774b5c879e589defca1e92,

title = "Bayesian group selection with non-local priors",

abstract = "In many applications, variables or features can be naturally partitioned into different groups. In this article, we propose a new Bayesian hierarchical model for group selection problem when the group structure is known. We use spike and slab priors for regression coefficients, and the slab component is assumed to come from the family of nonlocal priors. Contrary to local priors commonly used in Bayesian group selection, nonlocal density priors vanish when a regression coefficient in the model is zero. We use simulation studies to assess the performance of our method and apply it to data collected from individuals undergoing cardiac catheterization at Duke University Medical center between 2001 and 2010.",

keywords = "Bayesian computing, CATHGEN, MCMC, Spike and slab priors",

author = "Weibing Li and Thierry Chekouo",

note = "Funding Information: Thierry Chekouo is partially supported by NSERC Discovery Grants Number RGPIN-2019-04810. The content is solely the responsibility of the authors and does not necessarily represent the official views of NSERC. Publisher Copyright: {\textcopyright} 2021, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.",

year = "2022",

month = mar,

doi = "10.1007/s00180-021-01115-1",

language = "English (US)",

volume = "37",

pages = "287--302",

journal = "Computational Statistics",

issn = "0943-4062",

publisher = "Springer Verlag",

number = "1",

}

TY - JOUR

T1 - Bayesian group selection with non-local priors

AU - Li, Weibing

AU - Chekouo, Thierry

N1 - Funding Information: Thierry Chekouo is partially supported by NSERC Discovery Grants Number RGPIN-2019-04810. The content is solely the responsibility of the authors and does not necessarily represent the official views of NSERC. Publisher Copyright: © 2021, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.

PY - 2022/3

Y1 - 2022/3

N2 - In many applications, variables or features can be naturally partitioned into different groups. In this article, we propose a new Bayesian hierarchical model for group selection problem when the group structure is known. We use spike and slab priors for regression coefficients, and the slab component is assumed to come from the family of nonlocal priors. Contrary to local priors commonly used in Bayesian group selection, nonlocal density priors vanish when a regression coefficient in the model is zero. We use simulation studies to assess the performance of our method and apply it to data collected from individuals undergoing cardiac catheterization at Duke University Medical center between 2001 and 2010.

AB - In many applications, variables or features can be naturally partitioned into different groups. In this article, we propose a new Bayesian hierarchical model for group selection problem when the group structure is known. We use spike and slab priors for regression coefficients, and the slab component is assumed to come from the family of nonlocal priors. Contrary to local priors commonly used in Bayesian group selection, nonlocal density priors vanish when a regression coefficient in the model is zero. We use simulation studies to assess the performance of our method and apply it to data collected from individuals undergoing cardiac catheterization at Duke University Medical center between 2001 and 2010.

KW - Bayesian computing

KW - CATHGEN

KW - MCMC

KW - Spike and slab priors

UR - http://www.scopus.com/inward/record.url?scp=85106688656&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85106688656&partnerID=8YFLogxK

U2 - 10.1007/s00180-021-01115-1

DO - 10.1007/s00180-021-01115-1

M3 - Article

AN - SCOPUS:85106688656

SN - 0943-4062

VL - 37

SP - 287

EP - 302

JO - Computational Statistics

JF - Computational Statistics

IS - 1

ER -

Bayesian group selection with non-local priors

Abstract

Bibliographical note

Keywords

Access

OpenUrl availability

Other files and links

Fingerprint

Cite this