Measurement invariance, the lack thereof, and modeling change

Michael Edwards, Carrie R. Houts, R. J. Wirth

Research output: Contribution to journalArticle

6 Citations (Scopus)

Abstract

Purpose: Measurement invariance issues should be considered during test construction. In this paper, we provide a conceptual overview of measurement invariance and describe how the concept is implemented in several different statistical approaches. Typical applications look for invariance over things such as mode of administration (paper and pencil vs. computer based), language/translation, age, time, and gender, to cite just a few examples. To the extent that the relationships between items and constructs are stable/invariant, we can be more confident in score interpretations. Methods: A series of simulated examples are reported which highlight different kinds of non-invariance, the impact it can have, and the effect of appropriately modeling a lack of invariance. One example focuses on the longitudinal context, where measurement invariance is critical to understanding trends over time. Software syntax is provided to help researchers apply these models with their own data. Results: The simulation studies demonstrate the negative impact an erroneous assumption of invariance may have on scores and substantive conclusions drawn from naively analyzing those scores. Conclusions: Measurement invariance implies that the links between the items and the construct of interest are invariant over some domain, grouping, or classification. Examining a new or existing test for measurement invariance should be part of any test construction/implementation plan. In addition to reviewing implications of the simulation study results, we also provide a discussion of the limitations of current approaches and areas in need of additional research.

Original languageEnglish (US)
Pages (from-to)1735-1743
Number of pages9
JournalQuality of Life Research
Volume27
Issue number7
DOIs
StatePublished - Jul 1 2018

Fingerprint

Language
Software
Research Personnel
Research

Keywords

  • DIF
  • IRT
  • Measurement invariance
  • Psychometrics

ASJC Scopus subject areas

  • Public Health, Environmental and Occupational Health

Cite this

Measurement invariance, the lack thereof, and modeling change. / Edwards, Michael; Houts, Carrie R.; Wirth, R. J.

In: Quality of Life Research, Vol. 27, No. 7, 01.07.2018, p. 1735-1743.

Research output: Contribution to journalArticle

Edwards, Michael ; Houts, Carrie R. ; Wirth, R. J. / Measurement invariance, the lack thereof, and modeling change. In: Quality of Life Research. 2018 ; Vol. 27, No. 7. pp. 1735-1743.
@article{25119b7ab05243aaaf9da1ba892675c8,
title = "Measurement invariance, the lack thereof, and modeling change",
abstract = "Purpose: Measurement invariance issues should be considered during test construction. In this paper, we provide a conceptual overview of measurement invariance and describe how the concept is implemented in several different statistical approaches. Typical applications look for invariance over things such as mode of administration (paper and pencil vs. computer based), language/translation, age, time, and gender, to cite just a few examples. To the extent that the relationships between items and constructs are stable/invariant, we can be more confident in score interpretations. Methods: A series of simulated examples are reported which highlight different kinds of non-invariance, the impact it can have, and the effect of appropriately modeling a lack of invariance. One example focuses on the longitudinal context, where measurement invariance is critical to understanding trends over time. Software syntax is provided to help researchers apply these models with their own data. Results: The simulation studies demonstrate the negative impact an erroneous assumption of invariance may have on scores and substantive conclusions drawn from naively analyzing those scores. Conclusions: Measurement invariance implies that the links between the items and the construct of interest are invariant over some domain, grouping, or classification. Examining a new or existing test for measurement invariance should be part of any test construction/implementation plan. In addition to reviewing implications of the simulation study results, we also provide a discussion of the limitations of current approaches and areas in need of additional research.",
keywords = "DIF, IRT, Measurement invariance, Psychometrics",
author = "Michael Edwards and Houts, {Carrie R.} and Wirth, {R. J.}",
year = "2018",
month = "7",
day = "1",
doi = "10.1007/s11136-017-1673-7",
language = "English (US)",
volume = "27",
pages = "1735--1743",
journal = "Quality of Life Research",
issn = "0962-9343",
publisher = "Springer Netherlands",
number = "7",

}

TY - JOUR

T1 - Measurement invariance, the lack thereof, and modeling change

AU - Edwards, Michael

AU - Houts, Carrie R.

AU - Wirth, R. J.

PY - 2018/7/1

Y1 - 2018/7/1

N2 - Purpose: Measurement invariance issues should be considered during test construction. In this paper, we provide a conceptual overview of measurement invariance and describe how the concept is implemented in several different statistical approaches. Typical applications look for invariance over things such as mode of administration (paper and pencil vs. computer based), language/translation, age, time, and gender, to cite just a few examples. To the extent that the relationships between items and constructs are stable/invariant, we can be more confident in score interpretations. Methods: A series of simulated examples are reported which highlight different kinds of non-invariance, the impact it can have, and the effect of appropriately modeling a lack of invariance. One example focuses on the longitudinal context, where measurement invariance is critical to understanding trends over time. Software syntax is provided to help researchers apply these models with their own data. Results: The simulation studies demonstrate the negative impact an erroneous assumption of invariance may have on scores and substantive conclusions drawn from naively analyzing those scores. Conclusions: Measurement invariance implies that the links between the items and the construct of interest are invariant over some domain, grouping, or classification. Examining a new or existing test for measurement invariance should be part of any test construction/implementation plan. In addition to reviewing implications of the simulation study results, we also provide a discussion of the limitations of current approaches and areas in need of additional research.

AB - Purpose: Measurement invariance issues should be considered during test construction. In this paper, we provide a conceptual overview of measurement invariance and describe how the concept is implemented in several different statistical approaches. Typical applications look for invariance over things such as mode of administration (paper and pencil vs. computer based), language/translation, age, time, and gender, to cite just a few examples. To the extent that the relationships between items and constructs are stable/invariant, we can be more confident in score interpretations. Methods: A series of simulated examples are reported which highlight different kinds of non-invariance, the impact it can have, and the effect of appropriately modeling a lack of invariance. One example focuses on the longitudinal context, where measurement invariance is critical to understanding trends over time. Software syntax is provided to help researchers apply these models with their own data. Results: The simulation studies demonstrate the negative impact an erroneous assumption of invariance may have on scores and substantive conclusions drawn from naively analyzing those scores. Conclusions: Measurement invariance implies that the links between the items and the construct of interest are invariant over some domain, grouping, or classification. Examining a new or existing test for measurement invariance should be part of any test construction/implementation plan. In addition to reviewing implications of the simulation study results, we also provide a discussion of the limitations of current approaches and areas in need of additional research.

KW - DIF

KW - IRT

KW - Measurement invariance

KW - Psychometrics

UR - http://www.scopus.com/inward/record.url?scp=85027837227&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85027837227&partnerID=8YFLogxK

U2 - 10.1007/s11136-017-1673-7

DO - 10.1007/s11136-017-1673-7

M3 - Article

C2 - 28819711

AN - SCOPUS:85027837227

VL - 27

SP - 1735

EP - 1743

JO - Quality of Life Research

JF - Quality of Life Research

SN - 0962-9343

IS - 7

ER -