Robust and sparse gaussian graphical modelling under cell-wise contamination

Shota Katayama, Hironori Fujisawa, Mathias Drton

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Graphical modelling explores dependences among a collection of variables by inferring a graph that encodes pairwise conditional independences. For jointly Gaussian variables, this translates into detecting the support of the precision matrix. Many modern applications feature high-dimensional and contaminated data that complicate this task. In particular, traditional robust methods that down-weight entire observation vectors are often inappropriate as high-dimensional data may feature partial contamination in many observations. We tackle this problem by giving a robust method for sparse precision matrix estimation based on the γ-divergence under a cell-wise contamination model. Simulation studies demonstrate that our procedure outperforms existing methods especially for highly contaminated data.

Original languageEnglish
Article numbere181
JournalStat
Volume7
Issue number1
DOIs
Publication statusPublished - 2018 Jan 1
Externally publishedYes

Fingerprint

Graphical Modeling
Robust Methods
Contamination
Conditional Independence
Cell
High-dimensional Data
Pairwise
Divergence
High-dimensional
Simulation Study
Entire
Partial
Graph in graph theory
Demonstrate
Observation
Modeling
Model

Keywords

  • Cell-wise contamination
  • Gaussian graphical modelling
  • Precision matrix
  • Robust inference
  • Sparsity

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Cite this

Robust and sparse gaussian graphical modelling under cell-wise contamination. / Katayama, Shota; Fujisawa, Hironori; Drton, Mathias.

In: Stat, Vol. 7, No. 1, e181, 01.01.2018.

Research output: Contribution to journalArticle

Katayama, Shota ; Fujisawa, Hironori ; Drton, Mathias. / Robust and sparse gaussian graphical modelling under cell-wise contamination. In: Stat. 2018 ; Vol. 7, No. 1.
@article{f1c3b90c935047279acbaffa6d8e6199,
title = "Robust and sparse gaussian graphical modelling under cell-wise contamination",
abstract = "Graphical modelling explores dependences among a collection of variables by inferring a graph that encodes pairwise conditional independences. For jointly Gaussian variables, this translates into detecting the support of the precision matrix. Many modern applications feature high-dimensional and contaminated data that complicate this task. In particular, traditional robust methods that down-weight entire observation vectors are often inappropriate as high-dimensional data may feature partial contamination in many observations. We tackle this problem by giving a robust method for sparse precision matrix estimation based on the γ-divergence under a cell-wise contamination model. Simulation studies demonstrate that our procedure outperforms existing methods especially for highly contaminated data.",
keywords = "Cell-wise contamination, Gaussian graphical modelling, Precision matrix, Robust inference, Sparsity",
author = "Shota Katayama and Hironori Fujisawa and Mathias Drton",
year = "2018",
month = "1",
day = "1",
doi = "10.1002/sta4.181",
language = "English",
volume = "7",
journal = "Stat",
issn = "2049-1573",
publisher = "Wiley-Blackwell",
number = "1",

}

TY - JOUR

T1 - Robust and sparse gaussian graphical modelling under cell-wise contamination

AU - Katayama, Shota

AU - Fujisawa, Hironori

AU - Drton, Mathias

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Graphical modelling explores dependences among a collection of variables by inferring a graph that encodes pairwise conditional independences. For jointly Gaussian variables, this translates into detecting the support of the precision matrix. Many modern applications feature high-dimensional and contaminated data that complicate this task. In particular, traditional robust methods that down-weight entire observation vectors are often inappropriate as high-dimensional data may feature partial contamination in many observations. We tackle this problem by giving a robust method for sparse precision matrix estimation based on the γ-divergence under a cell-wise contamination model. Simulation studies demonstrate that our procedure outperforms existing methods especially for highly contaminated data.

AB - Graphical modelling explores dependences among a collection of variables by inferring a graph that encodes pairwise conditional independences. For jointly Gaussian variables, this translates into detecting the support of the precision matrix. Many modern applications feature high-dimensional and contaminated data that complicate this task. In particular, traditional robust methods that down-weight entire observation vectors are often inappropriate as high-dimensional data may feature partial contamination in many observations. We tackle this problem by giving a robust method for sparse precision matrix estimation based on the γ-divergence under a cell-wise contamination model. Simulation studies demonstrate that our procedure outperforms existing methods especially for highly contaminated data.

KW - Cell-wise contamination

KW - Gaussian graphical modelling

KW - Precision matrix

KW - Robust inference

KW - Sparsity

UR - http://www.scopus.com/inward/record.url?scp=85044327280&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85044327280&partnerID=8YFLogxK

U2 - 10.1002/sta4.181

DO - 10.1002/sta4.181

M3 - Article

AN - SCOPUS:85044327280

VL - 7

JO - Stat

JF - Stat

SN - 2049-1573

IS - 1

M1 - e181

ER -