High-dimensional multivariate mediation with application to neuroimaging data

Research output: Contribution to journalArticle

Abstract

Mediation analysis is an important tool in the behavioral sciences for investigating the role of intermediate variables that lie in the path between a treatment and an outcome variable. The influence of the intermediate variable on the outcome is often explored using a linear structural equation model (LSEM), with model coefficients interpreted as possible effects. While there has been significant research on the topic, little work has been done when the intermediate variable (mediator) is a high-dimensional vector. In this work, we introduce a novel method for identifying potential mediators in this setting called the directions of mediation (DMs). DMs linearly combine potential mediators into a smaller number of orthogonal components, with components ranked based on the proportion of the LSEM likelihood each accounts for. This method is well suited for cases when many potential mediators are measured. Examples of highdimensional potential mediators are brain images composed of hundreds of thousands of voxels, genetic variation measured at millions of single nucleotide polymorphisms (SNPs), or vectors of thousands of variables in large-scale epidemiological studies.We demonstrate the method using a functional magnetic resonance imaging study of thermal pain where we are interested in determining which brain locations mediate the relationship between the application of a thermal stimulus and self-reported pain.

Original languageEnglish (US)
Pages (from-to)121-136
Number of pages16
JournalBiostatistics
Volume19
Issue number2
DOIs
StatePublished - Apr 1 2018

Fingerprint

Neuroimaging
Mediation
Mediator
High-dimensional
Structural Equation Model
Pain
Linear equation
Genetic Variation
Functional Magnetic Resonance Imaging
Single nucleotide Polymorphism
Voxel
Likelihood
Proportion
Linearly
Path
Coefficient
Demonstrate

Keywords

  • Directions of mediation
  • fMRI, Mediation analysis
  • High-dimensional data
  • Principal components analysis
  • Structural equation models

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Cite this

High-dimensional multivariate mediation with application to neuroimaging data. / Chén, Oliver Y.; Crainiceanu, Ciprian M; Ogburn, Elizabeth Leigh; Caffo, Brian S; Wager, Tor D.; Lindquist, Martin.

In: Biostatistics, Vol. 19, No. 2, 01.04.2018, p. 121-136.

Research output: Contribution to journalArticle

@article{9499efd5a88f47c8bc0d3bbd52ddec9d,
title = "High-dimensional multivariate mediation with application to neuroimaging data",
abstract = "Mediation analysis is an important tool in the behavioral sciences for investigating the role of intermediate variables that lie in the path between a treatment and an outcome variable. The influence of the intermediate variable on the outcome is often explored using a linear structural equation model (LSEM), with model coefficients interpreted as possible effects. While there has been significant research on the topic, little work has been done when the intermediate variable (mediator) is a high-dimensional vector. In this work, we introduce a novel method for identifying potential mediators in this setting called the directions of mediation (DMs). DMs linearly combine potential mediators into a smaller number of orthogonal components, with components ranked based on the proportion of the LSEM likelihood each accounts for. This method is well suited for cases when many potential mediators are measured. Examples of highdimensional potential mediators are brain images composed of hundreds of thousands of voxels, genetic variation measured at millions of single nucleotide polymorphisms (SNPs), or vectors of thousands of variables in large-scale epidemiological studies.We demonstrate the method using a functional magnetic resonance imaging study of thermal pain where we are interested in determining which brain locations mediate the relationship between the application of a thermal stimulus and self-reported pain.",
keywords = "Directions of mediation, fMRI, Mediation analysis, High-dimensional data, Principal components analysis, Structural equation models",
author = "Ch{\'e}n, {Oliver Y.} and Crainiceanu, {Ciprian M} and Ogburn, {Elizabeth Leigh} and Caffo, {Brian S} and Wager, {Tor D.} and Martin Lindquist",
year = "2018",
month = "4",
day = "1",
doi = "10.1093/biostatistics/kxx027",
language = "English (US)",
volume = "19",
pages = "121--136",
journal = "Biostatistics",
issn = "1465-4644",
publisher = "Oxford University Press",
number = "2",

}

TY - JOUR

T1 - High-dimensional multivariate mediation with application to neuroimaging data

AU - Chén, Oliver Y.

AU - Crainiceanu, Ciprian M

AU - Ogburn, Elizabeth Leigh

AU - Caffo, Brian S

AU - Wager, Tor D.

AU - Lindquist, Martin

PY - 2018/4/1

Y1 - 2018/4/1

N2 - Mediation analysis is an important tool in the behavioral sciences for investigating the role of intermediate variables that lie in the path between a treatment and an outcome variable. The influence of the intermediate variable on the outcome is often explored using a linear structural equation model (LSEM), with model coefficients interpreted as possible effects. While there has been significant research on the topic, little work has been done when the intermediate variable (mediator) is a high-dimensional vector. In this work, we introduce a novel method for identifying potential mediators in this setting called the directions of mediation (DMs). DMs linearly combine potential mediators into a smaller number of orthogonal components, with components ranked based on the proportion of the LSEM likelihood each accounts for. This method is well suited for cases when many potential mediators are measured. Examples of highdimensional potential mediators are brain images composed of hundreds of thousands of voxels, genetic variation measured at millions of single nucleotide polymorphisms (SNPs), or vectors of thousands of variables in large-scale epidemiological studies.We demonstrate the method using a functional magnetic resonance imaging study of thermal pain where we are interested in determining which brain locations mediate the relationship between the application of a thermal stimulus and self-reported pain.

AB - Mediation analysis is an important tool in the behavioral sciences for investigating the role of intermediate variables that lie in the path between a treatment and an outcome variable. The influence of the intermediate variable on the outcome is often explored using a linear structural equation model (LSEM), with model coefficients interpreted as possible effects. While there has been significant research on the topic, little work has been done when the intermediate variable (mediator) is a high-dimensional vector. In this work, we introduce a novel method for identifying potential mediators in this setting called the directions of mediation (DMs). DMs linearly combine potential mediators into a smaller number of orthogonal components, with components ranked based on the proportion of the LSEM likelihood each accounts for. This method is well suited for cases when many potential mediators are measured. Examples of highdimensional potential mediators are brain images composed of hundreds of thousands of voxels, genetic variation measured at millions of single nucleotide polymorphisms (SNPs), or vectors of thousands of variables in large-scale epidemiological studies.We demonstrate the method using a functional magnetic resonance imaging study of thermal pain where we are interested in determining which brain locations mediate the relationship between the application of a thermal stimulus and self-reported pain.

KW - Directions of mediation

KW - fMRI, Mediation analysis

KW - High-dimensional data

KW - Principal components analysis

KW - Structural equation models

UR - http://www.scopus.com/inward/record.url?scp=85044742502&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85044742502&partnerID=8YFLogxK

U2 - 10.1093/biostatistics/kxx027

DO - 10.1093/biostatistics/kxx027

M3 - Article

C2 - 28637279

AN - SCOPUS:85044742502

VL - 19

SP - 121

EP - 136

JO - Biostatistics

JF - Biostatistics

SN - 1465-4644

IS - 2

ER -