Correcting Internal Consistency Estimates Inflated by Correlated Item Errors

Paola Pascual-Ferrá, Michael Beatty

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

Coefficient alpha is perhaps the most widely employed estimate of reliability for self-report measures in communication research. However, coefficient alpha misestimates internal consistency when measures are not congeneric. Congeneric measures meet the assumption of classical test theory in that all item errors are uncorrelated or random; when item errors are correlated, the measure fails to be congeneric, signaling the presence of latent factor(s) other than those the researcher intended to measure. Scholars have alerted to the fact that many self-report measures used are noncongeneric and that coefficient alpha might be yielding inaccurate, often inflated, reliability estimates. In this essay, we propose a formula for estimating the internal consistency of a measure that reflects the degree to which reliability is affected by correlated item errors.

Original languageEnglish (US)
Pages (from-to)347-352
Number of pages6
JournalCommunication Research Reports
Volume32
Issue number4
DOIs
StatePublished - Oct 2 2015

Fingerprint

test theory
communication research
Communication

Keywords

  • Coefficient Alpha
  • Congenericity
  • Correlated Item Errors
  • Measurement Error
  • Reliability

ASJC Scopus subject areas

  • Communication

Cite this

Correcting Internal Consistency Estimates Inflated by Correlated Item Errors. / Pascual-Ferrá, Paola; Beatty, Michael.

In: Communication Research Reports, Vol. 32, No. 4, 02.10.2015, p. 347-352.

Research output: Contribution to journalArticle

@article{34f9e5a9c98c4bc0aeda9816d8731126,
title = "Correcting Internal Consistency Estimates Inflated by Correlated Item Errors",
abstract = "Coefficient alpha is perhaps the most widely employed estimate of reliability for self-report measures in communication research. However, coefficient alpha misestimates internal consistency when measures are not congeneric. Congeneric measures meet the assumption of classical test theory in that all item errors are uncorrelated or random; when item errors are correlated, the measure fails to be congeneric, signaling the presence of latent factor(s) other than those the researcher intended to measure. Scholars have alerted to the fact that many self-report measures used are noncongeneric and that coefficient alpha might be yielding inaccurate, often inflated, reliability estimates. In this essay, we propose a formula for estimating the internal consistency of a measure that reflects the degree to which reliability is affected by correlated item errors.",
keywords = "Coefficient Alpha, Congenericity, Correlated Item Errors, Measurement Error, Reliability",
author = "Paola Pascual-Ferr{\'a} and Michael Beatty",
year = "2015",
month = "10",
day = "2",
doi = "10.1080/08824096.2015.1089858",
language = "English (US)",
volume = "32",
pages = "347--352",
journal = "Communication Research Reports",
issn = "0882-4096",
publisher = "Routledge",
number = "4",

}

TY - JOUR

T1 - Correcting Internal Consistency Estimates Inflated by Correlated Item Errors

AU - Pascual-Ferrá, Paola

AU - Beatty, Michael

PY - 2015/10/2

Y1 - 2015/10/2

N2 - Coefficient alpha is perhaps the most widely employed estimate of reliability for self-report measures in communication research. However, coefficient alpha misestimates internal consistency when measures are not congeneric. Congeneric measures meet the assumption of classical test theory in that all item errors are uncorrelated or random; when item errors are correlated, the measure fails to be congeneric, signaling the presence of latent factor(s) other than those the researcher intended to measure. Scholars have alerted to the fact that many self-report measures used are noncongeneric and that coefficient alpha might be yielding inaccurate, often inflated, reliability estimates. In this essay, we propose a formula for estimating the internal consistency of a measure that reflects the degree to which reliability is affected by correlated item errors.

AB - Coefficient alpha is perhaps the most widely employed estimate of reliability for self-report measures in communication research. However, coefficient alpha misestimates internal consistency when measures are not congeneric. Congeneric measures meet the assumption of classical test theory in that all item errors are uncorrelated or random; when item errors are correlated, the measure fails to be congeneric, signaling the presence of latent factor(s) other than those the researcher intended to measure. Scholars have alerted to the fact that many self-report measures used are noncongeneric and that coefficient alpha might be yielding inaccurate, often inflated, reliability estimates. In this essay, we propose a formula for estimating the internal consistency of a measure that reflects the degree to which reliability is affected by correlated item errors.

KW - Coefficient Alpha

KW - Congenericity

KW - Correlated Item Errors

KW - Measurement Error

KW - Reliability

UR - http://www.scopus.com/inward/record.url?scp=84944905859&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84944905859&partnerID=8YFLogxK

U2 - 10.1080/08824096.2015.1089858

DO - 10.1080/08824096.2015.1089858

M3 - Article

AN - SCOPUS:84944905859

VL - 32

SP - 347

EP - 352

JO - Communication Research Reports

JF - Communication Research Reports

SN - 0882-4096

IS - 4

ER -