Adding missing-data-relevant variables to FIML-based structural equation models

John W. Graham

Research output: Contribution to journalArticle

251 Citations (Scopus)

Abstract

Conventional wisdom in missing data research dictates adding variables to the missing data model when those variables are predictive of (a) missingness and (b) the variables containing missingness. However, it has recently been shown that adding variables that are correlated with variables containing missingness, whether or not they are related to missingness, can substantially improve estimation (bias and efficiency). Including large numbers of these "auxiliary" variables is straightforward for researchers who use multiple imputation. However, what is the researcher to do if 1 of the FIML/SEM procedures is the analysis of choice? This article suggests 2 models for SEM analysis with missing data, and presents simulation results to show that both models provide estimation that is clearly as good as analysis with the EM algorithm, and by extension, multiple imputation. One of these models, the saturated correlates model, also provides good estimates of model fit.

Original languageEnglish (US)
Pages (from-to)80-100
Number of pages21
JournalStructural Equation Modeling
Volume10
Issue number1
DOIs
StatePublished - Dec 1 2003

Fingerprint

Structural Equation Model
Missing Data
structural model
Multiple Imputation
Scanning electron microscopy
Model
Auxiliary Variables
EM Algorithm
Correlate
Data Model
Data structures
wisdom
Structural equation model
Missing data
efficiency
simulation
Estimate
trend
Simulation

All Science Journal Classification (ASJC) codes

  • Modeling and Simulation
  • Decision Sciences(all)
  • Economics, Econometrics and Finance(all)
  • Sociology and Political Science

Cite this

@article{4244beacba3e4610ac87206ca8651c23,
title = "Adding missing-data-relevant variables to FIML-based structural equation models",
abstract = "Conventional wisdom in missing data research dictates adding variables to the missing data model when those variables are predictive of (a) missingness and (b) the variables containing missingness. However, it has recently been shown that adding variables that are correlated with variables containing missingness, whether or not they are related to missingness, can substantially improve estimation (bias and efficiency). Including large numbers of these {"}auxiliary{"} variables is straightforward for researchers who use multiple imputation. However, what is the researcher to do if 1 of the FIML/SEM procedures is the analysis of choice? This article suggests 2 models for SEM analysis with missing data, and presents simulation results to show that both models provide estimation that is clearly as good as analysis with the EM algorithm, and by extension, multiple imputation. One of these models, the saturated correlates model, also provides good estimates of model fit.",
author = "Graham, {John W.}",
year = "2003",
month = "12",
day = "1",
doi = "10.1207/S15328007SEM1001_4",
language = "English (US)",
volume = "10",
pages = "80--100",
journal = "Structural Equation Modeling",
issn = "1070-5511",
publisher = "Psychology Press Ltd",
number = "1",

}

Adding missing-data-relevant variables to FIML-based structural equation models. / Graham, John W.

In: Structural Equation Modeling, Vol. 10, No. 1, 01.12.2003, p. 80-100.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Adding missing-data-relevant variables to FIML-based structural equation models

AU - Graham, John W.

PY - 2003/12/1

Y1 - 2003/12/1

N2 - Conventional wisdom in missing data research dictates adding variables to the missing data model when those variables are predictive of (a) missingness and (b) the variables containing missingness. However, it has recently been shown that adding variables that are correlated with variables containing missingness, whether or not they are related to missingness, can substantially improve estimation (bias and efficiency). Including large numbers of these "auxiliary" variables is straightforward for researchers who use multiple imputation. However, what is the researcher to do if 1 of the FIML/SEM procedures is the analysis of choice? This article suggests 2 models for SEM analysis with missing data, and presents simulation results to show that both models provide estimation that is clearly as good as analysis with the EM algorithm, and by extension, multiple imputation. One of these models, the saturated correlates model, also provides good estimates of model fit.

AB - Conventional wisdom in missing data research dictates adding variables to the missing data model when those variables are predictive of (a) missingness and (b) the variables containing missingness. However, it has recently been shown that adding variables that are correlated with variables containing missingness, whether or not they are related to missingness, can substantially improve estimation (bias and efficiency). Including large numbers of these "auxiliary" variables is straightforward for researchers who use multiple imputation. However, what is the researcher to do if 1 of the FIML/SEM procedures is the analysis of choice? This article suggests 2 models for SEM analysis with missing data, and presents simulation results to show that both models provide estimation that is clearly as good as analysis with the EM algorithm, and by extension, multiple imputation. One of these models, the saturated correlates model, also provides good estimates of model fit.

UR - http://www.scopus.com/inward/record.url?scp=0347249765&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0347249765&partnerID=8YFLogxK

U2 - 10.1207/S15328007SEM1001_4

DO - 10.1207/S15328007SEM1001_4

M3 - Article

VL - 10

SP - 80

EP - 100

JO - Structural Equation Modeling

JF - Structural Equation Modeling

SN - 1070-5511

IS - 1

ER -