Brief Research Report

Growth Models With Small Samples and Missing Data

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Small samples are common in growth models due to financial and logistical difficulties of following people longitudinally. For similar reasons, longitudinal studies often contain missing data. Though full information maximum likelihood (FIML) is popular to accommodate missing data, the limited number of studies in this area have found that FIML tends to perform poorly with small-sample growth models. This report demonstrates that the fault lies not with how FIML accommodates missingness but rather with maximum likelihood estimation itself. We discuss how the less popular restricted likelihood form of FIML, along with small-sample-appropriate methods, yields trustworthy estimates for growth models with small samples and missing data. That is, previously reported small sample issues with FIML are attributable to finite sample bias of maximum likelihood estimation not direct likelihood. Estimation issues pertinent to joint multiple imputation and predictive mean matching are also included and discussed.

Original languageEnglish (US)
Pages (from-to)1-12
Number of pages12
JournalJournal of Experimental Education
DOIs
StateAccepted/In press - Sep 15 2017

Fingerprint

Growth
Longitudinal Studies
Joints
longitudinal study
trend

Keywords

  • Growth model
  • missing data
  • mixed effect model
  • multilevel model
  • small sample

ASJC Scopus subject areas

  • Education
  • Developmental and Educational Psychology

Cite this

Brief Research Report : Growth Models With Small Samples and Missing Data. / McNeish, Daniel.

In: Journal of Experimental Education, 15.09.2017, p. 1-12.

Research output: Contribution to journalArticle

@article{2f79310075b34b62a2bc333bd2e176e2,
title = "Brief Research Report: Growth Models With Small Samples and Missing Data",
abstract = "Small samples are common in growth models due to financial and logistical difficulties of following people longitudinally. For similar reasons, longitudinal studies often contain missing data. Though full information maximum likelihood (FIML) is popular to accommodate missing data, the limited number of studies in this area have found that FIML tends to perform poorly with small-sample growth models. This report demonstrates that the fault lies not with how FIML accommodates missingness but rather with maximum likelihood estimation itself. We discuss how the less popular restricted likelihood form of FIML, along with small-sample-appropriate methods, yields trustworthy estimates for growth models with small samples and missing data. That is, previously reported small sample issues with FIML are attributable to finite sample bias of maximum likelihood estimation not direct likelihood. Estimation issues pertinent to joint multiple imputation and predictive mean matching are also included and discussed.",
keywords = "Growth model, missing data, mixed effect model, multilevel model, small sample",
author = "Daniel McNeish",
year = "2017",
month = "9",
day = "15",
doi = "10.1080/00220973.2017.1369384",
language = "English (US)",
pages = "1--12",
journal = "Journal of Experimental Education",
issn = "0022-0973",
publisher = "Routledge",

}

TY - JOUR

T1 - Brief Research Report

T2 - Growth Models With Small Samples and Missing Data

AU - McNeish, Daniel

PY - 2017/9/15

Y1 - 2017/9/15

N2 - Small samples are common in growth models due to financial and logistical difficulties of following people longitudinally. For similar reasons, longitudinal studies often contain missing data. Though full information maximum likelihood (FIML) is popular to accommodate missing data, the limited number of studies in this area have found that FIML tends to perform poorly with small-sample growth models. This report demonstrates that the fault lies not with how FIML accommodates missingness but rather with maximum likelihood estimation itself. We discuss how the less popular restricted likelihood form of FIML, along with small-sample-appropriate methods, yields trustworthy estimates for growth models with small samples and missing data. That is, previously reported small sample issues with FIML are attributable to finite sample bias of maximum likelihood estimation not direct likelihood. Estimation issues pertinent to joint multiple imputation and predictive mean matching are also included and discussed.

AB - Small samples are common in growth models due to financial and logistical difficulties of following people longitudinally. For similar reasons, longitudinal studies often contain missing data. Though full information maximum likelihood (FIML) is popular to accommodate missing data, the limited number of studies in this area have found that FIML tends to perform poorly with small-sample growth models. This report demonstrates that the fault lies not with how FIML accommodates missingness but rather with maximum likelihood estimation itself. We discuss how the less popular restricted likelihood form of FIML, along with small-sample-appropriate methods, yields trustworthy estimates for growth models with small samples and missing data. That is, previously reported small sample issues with FIML are attributable to finite sample bias of maximum likelihood estimation not direct likelihood. Estimation issues pertinent to joint multiple imputation and predictive mean matching are also included and discussed.

KW - Growth model

KW - missing data

KW - mixed effect model

KW - multilevel model

KW - small sample

UR - http://www.scopus.com/inward/record.url?scp=85029480087&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85029480087&partnerID=8YFLogxK

U2 - 10.1080/00220973.2017.1369384

DO - 10.1080/00220973.2017.1369384

M3 - Article

SP - 1

EP - 12

JO - Journal of Experimental Education

JF - Journal of Experimental Education

SN - 0022-0973

ER -