Outlier detection in contingency tables using decomposable graphical models

Mads Lindskou; Poul Svante Eriksen; Torben Tvedebrink

doi:10.1111/sjos.12407

Outlier detection in contingency tables using decomposable graphical models

Mads Lindskou^*, Poul Svante Eriksen, Torben Tvedebrink

^*Corresponding author for this work

Section of Forensic Genetics

Abstract

For high-dimensional data, it is a tedious task to determine anomalies such as outliers. We present a novel outlier detection method for high-dimensional contingency tables. We use the class of decomposable graphical models to model the relationship among the variables of interest, which can be depicted by an undirected graph called the interaction graph. Given an interaction graph, we derive a closed-form expression of the likelihood ratio test (LRT) statistic and an exact distribution for efficient simulation of the test statistic. An observation is declared an outlier if it deviates significantly from the approximated distribution of the test statistic under the null hypothesis. We demonstrate the use of the LRT outlier detection framework on genetic data modeled by Chow–Liu trees.

Original language	English
Journal	Scandinavian Journal of Statistics
ISSN	0303-6898
DOIs	https://doi.org/10.1111/sjos.12407
Publication status	Published - 1 Jun 2020

Keywords

categorical data
contingency table
decomposable
exact test
graphical model
likelihood ratio
outlier

Access to Document

10.1111/sjos.12407

Cite this

@article{f5097610b600434eba86bcd3bde83214,

title = "Outlier detection in contingency tables using decomposable graphical models",

abstract = "For high-dimensional data, it is a tedious task to determine anomalies such as outliers. We present a novel outlier detection method for high-dimensional contingency tables. We use the class of decomposable graphical models to model the relationship among the variables of interest, which can be depicted by an undirected graph called the interaction graph. Given an interaction graph, we derive a closed-form expression of the likelihood ratio test (LRT) statistic and an exact distribution for efficient simulation of the test statistic. An observation is declared an outlier if it deviates significantly from the approximated distribution of the test statistic under the null hypothesis. We demonstrate the use of the LRT outlier detection framework on genetic data modeled by Chow–Liu trees.",

keywords = "categorical data, contingency table, decomposable, exact test, graphical model, likelihood ratio, outlier",

author = "Mads Lindskou and Eriksen, {Poul Svante} and Torben Tvedebrink",

year = "2020",

month = jun,

day = "1",

doi = "10.1111/sjos.12407",

language = "English",

journal = "Scandinavian Journal of Statistics",

issn = "0303-6898",

publisher = "Wiley-Blackwell",

}

TY - JOUR

T1 - Outlier detection in contingency tables using decomposable graphical models

AU - Lindskou, Mads

AU - Eriksen, Poul Svante

AU - Tvedebrink, Torben

PY - 2020/6/1

Y1 - 2020/6/1

N2 - For high-dimensional data, it is a tedious task to determine anomalies such as outliers. We present a novel outlier detection method for high-dimensional contingency tables. We use the class of decomposable graphical models to model the relationship among the variables of interest, which can be depicted by an undirected graph called the interaction graph. Given an interaction graph, we derive a closed-form expression of the likelihood ratio test (LRT) statistic and an exact distribution for efficient simulation of the test statistic. An observation is declared an outlier if it deviates significantly from the approximated distribution of the test statistic under the null hypothesis. We demonstrate the use of the LRT outlier detection framework on genetic data modeled by Chow–Liu trees.

AB - For high-dimensional data, it is a tedious task to determine anomalies such as outliers. We present a novel outlier detection method for high-dimensional contingency tables. We use the class of decomposable graphical models to model the relationship among the variables of interest, which can be depicted by an undirected graph called the interaction graph. Given an interaction graph, we derive a closed-form expression of the likelihood ratio test (LRT) statistic and an exact distribution for efficient simulation of the test statistic. An observation is declared an outlier if it deviates significantly from the approximated distribution of the test statistic under the null hypothesis. We demonstrate the use of the LRT outlier detection framework on genetic data modeled by Chow–Liu trees.

KW - categorical data

KW - contingency table

KW - decomposable

KW - exact test

KW - graphical model

KW - likelihood ratio

KW - outlier

UR - http://www.scopus.com/inward/record.url?scp=85072173813&partnerID=8YFLogxK

U2 - 10.1111/sjos.12407

DO - 10.1111/sjos.12407

M3 - Journal article

AN - SCOPUS:85072173813

SN - 0303-6898

JO - Scandinavian Journal of Statistics

JF - Scandinavian Journal of Statistics

ER -

Outlier detection in contingency tables using decomposable graphical models

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this