Multilevel Matrix-Variate Analysis and its Application to Accelerometry-Measured Physical Activity in Clinical Populations

Lei Huang, Jiawei Bai, Andrada Ivanescu, Tamara Harris, Mathew Maurer, Philip Green, Vadim Zipunnikov

Research output: Contribution to journalArticleResearchpeer-review

Abstract

The number of studies where the primary measurement is a matrix is exploding. In response to this, we propose a statistical framework for modeling populations of repeatedly observed matrix-variate measurements. The 2D structure is handled via a matrix-variate distribution with decomposable row/column-specific covariance matrices and a linear mixed effect framework is used to model the multilevel design. The proposed framework flexibly expands to accommodate many common crossed and nested designs and introduces two important concepts: the between-subject distance and intraclass correlation coefficient, both defined for matrix-variate data. The computational feasibility and performance of the approach is shown in extensive simulation studies. The method is motivated by and applied to a study that monitored physical activity of individuals diagnosed with congestive heart failure (CHF) over a 4- to 9-month period. The long-term patterns of physical activity are studied and compared in two CHF subgroups: with and without adverse clinical events. Supplementary materials for this article, that include de-identified accelerometry and clinical data, are available online.

Original languageEnglish
Pages (from-to)553-564
Number of pages12
JournalJournal of the American Statistical Association
Volume114
Issue number526
DOIs
StatePublished - 3 Apr 2019

Fingerprint

Congestive Heart Failure
Nested Design
Intraclass Correlation Coefficient
Mixed Effects
Decomposable
Covariance matrix
Expand
Simulation Study
Subgroup
Physical activity
Modeling
Framework
Heart failure
Model
Concepts
Design
Correlation coefficient
Simulation study

Keywords

  • Actigraphy
  • Principal component analysis
  • Separable covariance

Cite this

Huang, Lei ; Bai, Jiawei ; Ivanescu, Andrada ; Harris, Tamara ; Maurer, Mathew ; Green, Philip ; Zipunnikov, Vadim. / Multilevel Matrix-Variate Analysis and its Application to Accelerometry-Measured Physical Activity in Clinical Populations. In: Journal of the American Statistical Association. 2019 ; Vol. 114, No. 526. pp. 553-564.
@article{7bb1162283134424b098dba03c65eaab,
title = "Multilevel Matrix-Variate Analysis and its Application to Accelerometry-Measured Physical Activity in Clinical Populations",
abstract = "The number of studies where the primary measurement is a matrix is exploding. In response to this, we propose a statistical framework for modeling populations of repeatedly observed matrix-variate measurements. The 2D structure is handled via a matrix-variate distribution with decomposable row/column-specific covariance matrices and a linear mixed effect framework is used to model the multilevel design. The proposed framework flexibly expands to accommodate many common crossed and nested designs and introduces two important concepts: the between-subject distance and intraclass correlation coefficient, both defined for matrix-variate data. The computational feasibility and performance of the approach is shown in extensive simulation studies. The method is motivated by and applied to a study that monitored physical activity of individuals diagnosed with congestive heart failure (CHF) over a 4- to 9-month period. The long-term patterns of physical activity are studied and compared in two CHF subgroups: with and without adverse clinical events. Supplementary materials for this article, that include de-identified accelerometry and clinical data, are available online.",
keywords = "Actigraphy, Principal component analysis, Separable covariance",
author = "Lei Huang and Jiawei Bai and Andrada Ivanescu and Tamara Harris and Mathew Maurer and Philip Green and Vadim Zipunnikov",
year = "2019",
month = "4",
day = "3",
doi = "10.1080/01621459.2018.1482750",
language = "English",
volume = "114",
pages = "553--564",
journal = "Journal of the American Statistical Association",
issn = "0162-1459",
publisher = "Taylor and Francis Ltd.",
number = "526",

}

Multilevel Matrix-Variate Analysis and its Application to Accelerometry-Measured Physical Activity in Clinical Populations. / Huang, Lei; Bai, Jiawei; Ivanescu, Andrada; Harris, Tamara; Maurer, Mathew; Green, Philip; Zipunnikov, Vadim.

In: Journal of the American Statistical Association, Vol. 114, No. 526, 03.04.2019, p. 553-564.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Multilevel Matrix-Variate Analysis and its Application to Accelerometry-Measured Physical Activity in Clinical Populations

AU - Huang, Lei

AU - Bai, Jiawei

AU - Ivanescu, Andrada

AU - Harris, Tamara

AU - Maurer, Mathew

AU - Green, Philip

AU - Zipunnikov, Vadim

PY - 2019/4/3

Y1 - 2019/4/3

N2 - The number of studies where the primary measurement is a matrix is exploding. In response to this, we propose a statistical framework for modeling populations of repeatedly observed matrix-variate measurements. The 2D structure is handled via a matrix-variate distribution with decomposable row/column-specific covariance matrices and a linear mixed effect framework is used to model the multilevel design. The proposed framework flexibly expands to accommodate many common crossed and nested designs and introduces two important concepts: the between-subject distance and intraclass correlation coefficient, both defined for matrix-variate data. The computational feasibility and performance of the approach is shown in extensive simulation studies. The method is motivated by and applied to a study that monitored physical activity of individuals diagnosed with congestive heart failure (CHF) over a 4- to 9-month period. The long-term patterns of physical activity are studied and compared in two CHF subgroups: with and without adverse clinical events. Supplementary materials for this article, that include de-identified accelerometry and clinical data, are available online.

AB - The number of studies where the primary measurement is a matrix is exploding. In response to this, we propose a statistical framework for modeling populations of repeatedly observed matrix-variate measurements. The 2D structure is handled via a matrix-variate distribution with decomposable row/column-specific covariance matrices and a linear mixed effect framework is used to model the multilevel design. The proposed framework flexibly expands to accommodate many common crossed and nested designs and introduces two important concepts: the between-subject distance and intraclass correlation coefficient, both defined for matrix-variate data. The computational feasibility and performance of the approach is shown in extensive simulation studies. The method is motivated by and applied to a study that monitored physical activity of individuals diagnosed with congestive heart failure (CHF) over a 4- to 9-month period. The long-term patterns of physical activity are studied and compared in two CHF subgroups: with and without adverse clinical events. Supplementary materials for this article, that include de-identified accelerometry and clinical data, are available online.

KW - Actigraphy

KW - Principal component analysis

KW - Separable covariance

UR - http://www.scopus.com/inward/record.url?scp=85052119044&partnerID=8YFLogxK

U2 - 10.1080/01621459.2018.1482750

DO - 10.1080/01621459.2018.1482750

M3 - Article

VL - 114

SP - 553

EP - 564

JO - Journal of the American Statistical Association

JF - Journal of the American Statistical Association

SN - 0162-1459

IS - 526

ER -