The impact of misspecifying the within-subject covariance structure in multiwave longitudinal multilevel models: A Monte Carlo study

Oi Man Kwok, Stephen West, Samuel B. Green

Research output: Contribution to journalArticle

66 Citations (Scopus)

Abstract

This Monte Carlo study examined the impact of misspecifying the σ matrix in longitudinal data analysis under both the multilevel model and mixed model frameworks. Under the multilevel model approach, under-specification and generalmisspecification of the σ matrix usually resulted in overestimation of the variances of the random effects (e.g., τ00, τ τ11) and standard errors of the corresponding growth parameter estimates (e.g., SE β0, SE β1). Overestimates of the standard errors led to lower statistical power in tests of the growth parameters. An unstructured σ matrix under the mixed model framework generally led to underestimates of standard errors of the growth parameter estimates. Underestimates of the standard errors led to inflation of me type I error rate in tests of the growth parameters. Implications of the compensatory relationship between the random effects of the growth parameters and the longitudinal error structure for model specification were discussed.

Original languageEnglish (US)
Pages (from-to)557-592
Number of pages36
JournalMultivariate Behavioral Research
Volume42
Issue number3
StatePublished - 2007

Fingerprint

Multilevel Models
Covariance Structure
Monte Carlo Study
Standard error
Growth
Mixed Model
Random Effects
Longitudinal Data Analysis
Statistical Power
Type I Error Rate
Model Specification
Economic Inflation
Inflation
Estimate
inflation
Specification
data analysis

ASJC Scopus subject areas

  • Mathematics (miscellaneous)
  • Statistics and Probability
  • Social Sciences (miscellaneous)
  • Psychology(all)
  • Experimental and Cognitive Psychology

Cite this

The impact of misspecifying the within-subject covariance structure in multiwave longitudinal multilevel models : A Monte Carlo study. / Kwok, Oi Man; West, Stephen; Green, Samuel B.

In: Multivariate Behavioral Research, Vol. 42, No. 3, 2007, p. 557-592.

Research output: Contribution to journalArticle

@article{79b8d4f8bd4a4c8e91d0afc5b7ea65bd,
title = "The impact of misspecifying the within-subject covariance structure in multiwave longitudinal multilevel models: A Monte Carlo study",
abstract = "This Monte Carlo study examined the impact of misspecifying the σ matrix in longitudinal data analysis under both the multilevel model and mixed model frameworks. Under the multilevel model approach, under-specification and generalmisspecification of the σ matrix usually resulted in overestimation of the variances of the random effects (e.g., τ00, τ τ11) and standard errors of the corresponding growth parameter estimates (e.g., SE β0, SE β1). Overestimates of the standard errors led to lower statistical power in tests of the growth parameters. An unstructured σ matrix under the mixed model framework generally led to underestimates of standard errors of the growth parameter estimates. Underestimates of the standard errors led to inflation of me type I error rate in tests of the growth parameters. Implications of the compensatory relationship between the random effects of the growth parameters and the longitudinal error structure for model specification were discussed.",
author = "Kwok, {Oi Man} and Stephen West and Green, {Samuel B.}",
year = "2007",
language = "English (US)",
volume = "42",
pages = "557--592",
journal = "Multivariate Behavioral Research",
issn = "0027-3171",
publisher = "Psychology Press Ltd",
number = "3",

}

TY - JOUR

T1 - The impact of misspecifying the within-subject covariance structure in multiwave longitudinal multilevel models

T2 - A Monte Carlo study

AU - Kwok, Oi Man

AU - West, Stephen

AU - Green, Samuel B.

PY - 2007

Y1 - 2007

N2 - This Monte Carlo study examined the impact of misspecifying the σ matrix in longitudinal data analysis under both the multilevel model and mixed model frameworks. Under the multilevel model approach, under-specification and generalmisspecification of the σ matrix usually resulted in overestimation of the variances of the random effects (e.g., τ00, τ τ11) and standard errors of the corresponding growth parameter estimates (e.g., SE β0, SE β1). Overestimates of the standard errors led to lower statistical power in tests of the growth parameters. An unstructured σ matrix under the mixed model framework generally led to underestimates of standard errors of the growth parameter estimates. Underestimates of the standard errors led to inflation of me type I error rate in tests of the growth parameters. Implications of the compensatory relationship between the random effects of the growth parameters and the longitudinal error structure for model specification were discussed.

AB - This Monte Carlo study examined the impact of misspecifying the σ matrix in longitudinal data analysis under both the multilevel model and mixed model frameworks. Under the multilevel model approach, under-specification and generalmisspecification of the σ matrix usually resulted in overestimation of the variances of the random effects (e.g., τ00, τ τ11) and standard errors of the corresponding growth parameter estimates (e.g., SE β0, SE β1). Overestimates of the standard errors led to lower statistical power in tests of the growth parameters. An unstructured σ matrix under the mixed model framework generally led to underestimates of standard errors of the growth parameter estimates. Underestimates of the standard errors led to inflation of me type I error rate in tests of the growth parameters. Implications of the compensatory relationship between the random effects of the growth parameters and the longitudinal error structure for model specification were discussed.

UR - http://www.scopus.com/inward/record.url?scp=36048985795&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=36048985795&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:36048985795

VL - 42

SP - 557

EP - 592

JO - Multivariate Behavioral Research

JF - Multivariate Behavioral Research

SN - 0027-3171

IS - 3

ER -