Mining exceptional mediation models

Florian Lemmerich; Christoph Kiefer; Benedikt Langenberg; Jeffry Cacho Aboukhalil; Axel Mayer

doi:10.1007/978-3-030-59491-6_30

Mining exceptional mediation models

Florian Lemmerich^*
, Christoph Kiefer
, Benedikt Langenberg
, Jeffry Cacho Aboukhalil
, Axel Mayer

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceeding › Academic › peer-review

Abstract

In statistics, mediation models aim to identify and explain the direct and indirect effects of an independent variable on a dependent variable. In heterogeneous data, the observed effects might vary for parts of the data. In this paper, we develop an approach for identifying interpretable data subgroups that induce exceptionally different effects in a mediation model. For that purpose, we introduce mediation models as a novel model class for the exceptional model mining framework, introduce suitable interestingness measures for several subtasks, and demonstrate the benefits of our approach on synthetic and empirical datasets.

Original language	English
Title of host publication	Foundations Of Intelligent Systems (ismis 2020)
Editors	D Helic, G Leitner, M Stettinger, A Felfernig, ZW Ras
Place of Publication	Graz, Austria
Publisher	Springer International Publishing
Pages	318-328
Number of pages	11
Volume	12117
ISBN (Electronic)	978-3-030-59491-6
ISBN (Print)	978-3-030-59490-9
DOIs	https://doi.org/10.1007/978-3-030-59491-6_30
Publication status	Published - 2020
Externally published	Yes

Access to Document

10.1007/978-3-030-59491-6_30

Cite this

@inproceedings{c57a34df10454c0e8a52e3584fc7b4e3,

title = "Mining exceptional mediation models",

abstract = "In statistics, mediation models aim to identify and explain the direct and indirect effects of an independent variable on a dependent variable. In heterogeneous data, the observed effects might vary for parts of the data. In this paper, we develop an approach for identifying interpretable data subgroups that induce exceptionally different effects in a mediation model. For that purpose, we introduce mediation models as a novel model class for the exceptional model mining framework, introduce suitable interestingness measures for several subtasks, and demonstrate the benefits of our approach on synthetic and empirical datasets.",

author = "Florian Lemmerich and Christoph Kiefer and Benedikt Langenberg and \{Cacho Aboukhalil\}, Jeffry and Axel Mayer",

year = "2020",

doi = "10.1007/978-3-030-59491-6\_30",

language = "English",

isbn = "978-3-030-59490-9",

volume = "12117",

pages = "318--328",

editor = "D Helic and G Leitner and M Stettinger and A Felfernig and ZW Ras",

booktitle = "Foundations Of Intelligent Systems (ismis 2020)",

publisher = "Springer International Publishing",

address = "Switzerland",

}

Mining exceptional mediation models. / Lemmerich, Florian; Kiefer, Christoph; Langenberg, Benedikt et al.
Foundations Of Intelligent Systems (ismis 2020). ed. / D Helic; G Leitner; M Stettinger; A Felfernig; ZW Ras. Vol. 12117 Graz, Austria: Springer International Publishing, 2020. p. 318-328.

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceeding › Academic › peer-review

TY - GEN

T1 - Mining exceptional mediation models

AU - Lemmerich, Florian

AU - Kiefer, Christoph

AU - Langenberg, Benedikt

AU - Cacho Aboukhalil, Jeffry

AU - Mayer, Axel

PY - 2020

Y1 - 2020

N2 - In statistics, mediation models aim to identify and explain the direct and indirect effects of an independent variable on a dependent variable. In heterogeneous data, the observed effects might vary for parts of the data. In this paper, we develop an approach for identifying interpretable data subgroups that induce exceptionally different effects in a mediation model. For that purpose, we introduce mediation models as a novel model class for the exceptional model mining framework, introduce suitable interestingness measures for several subtasks, and demonstrate the benefits of our approach on synthetic and empirical datasets.

AB - In statistics, mediation models aim to identify and explain the direct and indirect effects of an independent variable on a dependent variable. In heterogeneous data, the observed effects might vary for parts of the data. In this paper, we develop an approach for identifying interpretable data subgroups that induce exceptionally different effects in a mediation model. For that purpose, we introduce mediation models as a novel model class for the exceptional model mining framework, introduce suitable interestingness measures for several subtasks, and demonstrate the benefits of our approach on synthetic and empirical datasets.

U2 - 10.1007/978-3-030-59491-6_30

DO - 10.1007/978-3-030-59491-6_30

M3 - Conference article in proceeding

SN - 978-3-030-59490-9

VL - 12117

SP - 318

EP - 328

BT - Foundations Of Intelligent Systems (ismis 2020)

A2 - Helic, D

A2 - Leitner, G

A2 - Stettinger, M

A2 - Felfernig, A

A2 - Ras, ZW

PB - Springer International Publishing

CY - Graz, Austria

ER -

Mining exceptional mediation models

Abstract

Access to Document

Fingerprint

Cite this