Modeling time-varying effects with generalized and unsynchronized longitudinal data

Damla Şentürk, Lorien Dalrymple, Sandra M. Mohammed, George Kaysen, Danh V. Nguyen

Research output: Contribution to journalArticle

15 Citations (Scopus)

Abstract

We propose novel estimation approaches for generalized varying coefficient models that are tailored for unsynchronized, irregular and infrequent longitudinal designs/data. Unsynchronized longitudinal data refer to the time-dependent response and covariate measurements for each individual measured at distinct time points. Data from the Comprehensive Dialysis Study motivate the proposed methods. We model the potential age-varying association between infection-related hospitalization status and the inflammatory marker, C-reactive protein, within the first 2years from initiation of dialysis. We cannot directly apply traditional longitudinal modeling to unsynchronized data, and no method exists to estimate time-varying or age-varying effects for generalized outcomes (e.g., binary or count data) to date. In addition, through the analysis of the Comprehensive Dialysis Study data and simulation studies, we show that preprocessing steps, such as binning, needed to synchronize data to apply traditional modeling can lead to significant loss of information in this context. In contrast, the proposed approaches discard no observation; they exploit the fact that although there is little information in a single subject trajectory because of irregularity and infrequency, the moments of the underlying processes can be accurately and efficiently recovered by pooling information from all subjects using functional data analysis. We derive subject-specific mean response trajectory predictions and study finite sample properties of the estimators.

Original languageEnglish (US)
Pages (from-to)2971-2987
Number of pages17
JournalStatistics in Medicine
Volume32
Issue number17
DOIs
StatePublished - Jul 30 2013

Fingerprint

Longitudinal Data
Dialysis
Time-varying
Modeling
C-Reactive Protein
Trajectory
Hospitalization
Functional Data Analysis
Varying Coefficient Model
Binning
Observation
Pooling
Count Data
Binary Data
Irregularity
Infection
Preprocessing
Covariates
Irregular
Simulation Study

Keywords

  • Binning
  • Functional data analysis
  • Generalized linear models
  • Sparse design
  • United States Renal Data System
  • Varying coefficient models

ASJC Scopus subject areas

  • Epidemiology
  • Statistics and Probability

Cite this

Modeling time-varying effects with generalized and unsynchronized longitudinal data. / Şentürk, Damla; Dalrymple, Lorien; Mohammed, Sandra M.; Kaysen, George; Nguyen, Danh V.

In: Statistics in Medicine, Vol. 32, No. 17, 30.07.2013, p. 2971-2987.

Research output: Contribution to journalArticle

Şentürk, Damla ; Dalrymple, Lorien ; Mohammed, Sandra M. ; Kaysen, George ; Nguyen, Danh V. / Modeling time-varying effects with generalized and unsynchronized longitudinal data. In: Statistics in Medicine. 2013 ; Vol. 32, No. 17. pp. 2971-2987.
@article{11674ab4bd51498fba44839a6d2cd870,
title = "Modeling time-varying effects with generalized and unsynchronized longitudinal data",
abstract = "We propose novel estimation approaches for generalized varying coefficient models that are tailored for unsynchronized, irregular and infrequent longitudinal designs/data. Unsynchronized longitudinal data refer to the time-dependent response and covariate measurements for each individual measured at distinct time points. Data from the Comprehensive Dialysis Study motivate the proposed methods. We model the potential age-varying association between infection-related hospitalization status and the inflammatory marker, C-reactive protein, within the first 2years from initiation of dialysis. We cannot directly apply traditional longitudinal modeling to unsynchronized data, and no method exists to estimate time-varying or age-varying effects for generalized outcomes (e.g., binary or count data) to date. In addition, through the analysis of the Comprehensive Dialysis Study data and simulation studies, we show that preprocessing steps, such as binning, needed to synchronize data to apply traditional modeling can lead to significant loss of information in this context. In contrast, the proposed approaches discard no observation; they exploit the fact that although there is little information in a single subject trajectory because of irregularity and infrequency, the moments of the underlying processes can be accurately and efficiently recovered by pooling information from all subjects using functional data analysis. We derive subject-specific mean response trajectory predictions and study finite sample properties of the estimators.",
keywords = "Binning, Functional data analysis, Generalized linear models, Sparse design, United States Renal Data System, Varying coefficient models",
author = "Damla Şent{\"u}rk and Lorien Dalrymple and Mohammed, {Sandra M.} and George Kaysen and Nguyen, {Danh V.}",
year = "2013",
month = "7",
day = "30",
doi = "10.1002/sim.5740",
language = "English (US)",
volume = "32",
pages = "2971--2987",
journal = "Statistics in Medicine",
issn = "0277-6715",
publisher = "John Wiley and Sons Ltd",
number = "17",

}

TY - JOUR

T1 - Modeling time-varying effects with generalized and unsynchronized longitudinal data

AU - Şentürk, Damla

AU - Dalrymple, Lorien

AU - Mohammed, Sandra M.

AU - Kaysen, George

AU - Nguyen, Danh V.

PY - 2013/7/30

Y1 - 2013/7/30

N2 - We propose novel estimation approaches for generalized varying coefficient models that are tailored for unsynchronized, irregular and infrequent longitudinal designs/data. Unsynchronized longitudinal data refer to the time-dependent response and covariate measurements for each individual measured at distinct time points. Data from the Comprehensive Dialysis Study motivate the proposed methods. We model the potential age-varying association between infection-related hospitalization status and the inflammatory marker, C-reactive protein, within the first 2years from initiation of dialysis. We cannot directly apply traditional longitudinal modeling to unsynchronized data, and no method exists to estimate time-varying or age-varying effects for generalized outcomes (e.g., binary or count data) to date. In addition, through the analysis of the Comprehensive Dialysis Study data and simulation studies, we show that preprocessing steps, such as binning, needed to synchronize data to apply traditional modeling can lead to significant loss of information in this context. In contrast, the proposed approaches discard no observation; they exploit the fact that although there is little information in a single subject trajectory because of irregularity and infrequency, the moments of the underlying processes can be accurately and efficiently recovered by pooling information from all subjects using functional data analysis. We derive subject-specific mean response trajectory predictions and study finite sample properties of the estimators.

AB - We propose novel estimation approaches for generalized varying coefficient models that are tailored for unsynchronized, irregular and infrequent longitudinal designs/data. Unsynchronized longitudinal data refer to the time-dependent response and covariate measurements for each individual measured at distinct time points. Data from the Comprehensive Dialysis Study motivate the proposed methods. We model the potential age-varying association between infection-related hospitalization status and the inflammatory marker, C-reactive protein, within the first 2years from initiation of dialysis. We cannot directly apply traditional longitudinal modeling to unsynchronized data, and no method exists to estimate time-varying or age-varying effects for generalized outcomes (e.g., binary or count data) to date. In addition, through the analysis of the Comprehensive Dialysis Study data and simulation studies, we show that preprocessing steps, such as binning, needed to synchronize data to apply traditional modeling can lead to significant loss of information in this context. In contrast, the proposed approaches discard no observation; they exploit the fact that although there is little information in a single subject trajectory because of irregularity and infrequency, the moments of the underlying processes can be accurately and efficiently recovered by pooling information from all subjects using functional data analysis. We derive subject-specific mean response trajectory predictions and study finite sample properties of the estimators.

KW - Binning

KW - Functional data analysis

KW - Generalized linear models

KW - Sparse design

KW - United States Renal Data System

KW - Varying coefficient models

UR - http://www.scopus.com/inward/record.url?scp=84880041246&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84880041246&partnerID=8YFLogxK

U2 - 10.1002/sim.5740

DO - 10.1002/sim.5740

M3 - Article

C2 - 23335196

AN - SCOPUS:84880041246

VL - 32

SP - 2971

EP - 2987

JO - Statistics in Medicine

JF - Statistics in Medicine

SN - 0277-6715

IS - 17

ER -