VerifiedEvidence: highv1.0.0

Discriminant Validity

A form of construct validity shown when an instrument's scores do not correlate strongly with measures of theoretically distinct, unrelated concepts.

Last reviewedDarrin Baines IP Ltd

Concept Architecture

Concept

Theoretically, Discriminant Validity is a component of construct validity that assesses whether a measurement instrument is sufficiently distinct from instruments measuring different theoretical constructs. It is founded on psychometric theory and the multitrait-multimethod framework, which predicts that measures of unrelated constructs should demonstrate weak associations. Discriminant validity exists to ensure that an instrument captures its intended construct without inadvertently measuring different concepts.

Mathematically, discriminant validity is evaluated by examining the magnitude of relationships between theoretically unrelated measures. Correlation coefficients, confirmatory factor analysis and structural equation modelling are commonly used to assess whether constructs remain statistically distinguishable. Modern psychometric validation frequently evaluates discriminant validity using the Fornell-Larcker criterion or the heterotrait-monotrait ratio of correlations.

In practice, discriminant validity is assessed during the validation of clinical outcome assessments, patient-reported outcome measures and health-related quality-of-life instruments by comparing them with measures representing different constructs. Health economists evaluate discriminant validity to ensure that utility instruments and economic outcome measures distinguish between separate health domains and avoid construct overlap.


Purpose

Used to determine whether a measurement instrument measures a unique construct that is distinct from other theoretical concepts, supporting the validity of health outcome measures used in research and economic evaluation.


Mathematical Formulae

Primary Formula

There is no universally recognised canonical mathematical formula.

Supporting Formulae

Pearson correlation coefficient:

r = Cov(X,Y) / (�?�?)

Fornell-Larcker criterion:

AVE > r�

where:

AVE = Average Variance Extracted

r = correlation between constructs

Heterotrait-Monotrait Ratio:

HTMT = Mean(heterotrait correlations) � Mean(monotrait correlations)

Related Mathematical Methods

Construct Validity

Convergent Validity

Confirmatory Factor Analysis

Structural Equation Modelling

Average Variance Extracted

Heterotrait-Monotrait Ratio

Multitrait-Multimethod Analysis


Example

A health-related quality-of-life questionnaire is expected to measure physical functioning rather than emotional wellbeing. Validation testing demonstrates a weak correlation (r = 0.18) with an anxiety scale but a strong correlation with an established physical functioning measure. These findings indicate that the questionnaire measures a distinct construct and therefore demonstrates good discriminant validity.


Excel Implementation

FunctionExample FormulaHealth Economics Application
CORREL=CORREL(B2:B101,C2:C101)Assess correlations between theoretically unrelated constructs
RSQ=RSQ(B2:B101,C2:C101)Compare shared variance between constructs
AVERAGE=AVERAGE(B2:B101)Summarise construct scores
STDEV.S=STDEV.S(B2:B101)Assess variability during validation
IF=IF(ABS(CORREL(B2:B101,C2:C101))<0.30,"Supported","Review")Evaluate evidence supporting discriminant validity

VBA (Optional)

VBA can automate discriminant validity analyses by calculating correlation matrices, Fornell-Larcker comparisons and psychometric validation reports across multiple measurement instruments.


Sources

  • Campbell DT, Fiske DW. Convergent and Discriminant Validation by the Multitrait-Multimethod Matrix.
  • Fornell C, Larcker DF. Evaluating Structural Equation Models with Unobservable Variables and Measurement Error.
  • Henseler J, Ringle CM, Sarstedt M. A New Criterion for Assessing Discriminant Validity in Variance-Based Structural Equation Modelling.
  • Streiner DL, Norman GR, Cairney J. Health Measurement Scales: A Practical Guide to Their Development and Use.
  • COSMIN Initiative. COSMIN Methodology for Evaluating Measurement Properties.

Library

Publications

1
  • Journal article

    GRADE Guidelines: 1. Introduction — GRADE Evidence Profiles and Summary of Findings Tables — Guyatt, Oxman, Akl, Kunz, Vist, Brozek, et al., GRADE Series ed., 2011 (Journal of Clinical Epidemiology)

    The introductory paper of the GRADE (Grading of Recommendations Assessment, Development and Evaluation) series, setting out how to rate certainty of evidence (high/moderate/low/very low) and build evidence profiles and summary-of-findings tables.

Frequently Asked Questions (6)

  • What is discriminant validity?

    A form of construct validity shown when an instrument's scores do not correlate strongly with measures of theoretically distinct, unrelated concepts.

    Source: Campbell & Fiske 1959

  • What does discriminant validity expect of unrelated measures?

    Discriminant validity expects that an instrument's scores will not correlate strongly with measures of concepts it is meant to be distinct from, showing that it picks out its own target rather than a general tendency. A measure of anxiety, for instance, should not track a measure of physical mobility closely, since the two are conceptually separate. Weak correlation with such unrelated measures confirms the instrument discriminates between different things. Separation from the unrelated is its test. Streiner and Norman (2008) describe this form of validity.

    Source: Streiner & Norman 2008

  • How is discriminant validity assessed?

    Discriminant validity is assessed by correlating an instrument's scores with measures of concepts that are theoretically distinct and should not be closely related, and examining whether the correlations are weak, as expected. A low correlation with an unrelated measure supports discriminant validity, while an unexpectedly high correlation suggests the instrument may not be capturing a distinct construct. The comparison measures should represent genuinely different concepts. So discriminant validity is demonstrated by showing that the instrument's scores do not correlate strongly with measures of unrelated concepts, providing evidence that it measures a specific, distinct construct.

    Source: Campbell & Fiske 1959

  • Why is discriminant validity important?

    Discriminant validity is important because it provides evidence that an instrument measures a specific, distinct construct rather than overlapping with unrelated concepts, complementing convergent validity in establishing construct validity. Without discriminant validity, an instrument might correlate with many things, suggesting it captures a broad or ill-defined concept rather than the intended one. Demonstrating that scores differ from unrelated measures, as they should, supports the specificity of the instrument. So discriminant validity helps confirm that an instrument's scores meaningfully reflect the particular construct of interest and not a different or broader concept.

    Source: Campbell & Fiske 1959

  • How does discriminant validity complement convergent validity?

    Discriminant validity complements convergent validity by addressing the other side of construct validation: convergent validity shows that scores correlate with measures of related concepts, as expected, while discriminant validity shows that they do not correlate strongly with measures of unrelated concepts. Together they demonstrate that the instrument agrees with what it should and differs from what it should, providing a fuller test of whether it captures the intended construct specifically. Campbell and Fiske proposed assessing both through the multitrait-multimethod matrix. So the two are jointly needed for convincing evidence of construct validity.

    Source: Campbell & Fiske 1959

  • What does poor discriminant validity indicate?

    Poor discriminant validity, indicated by an instrument's scores correlating strongly with measures of concepts that should be unrelated, suggests that the instrument may not be capturing a distinct construct but instead overlapping with other concepts, or that the constructs are not as distinct as assumed, or that a shared method or bias inflates the correlation. It raises doubt about the specificity of what the instrument measures. So poor discriminant validity signals a problem with construct validity, prompting examination of the instrument, the constructs, or the measurement, since the instrument does not clearly distinguish the intended concept from others.

    Source: Campbell & Fiske 1959

Trust Record

Verified by Dr Darrin Baines

British health economist

Professional identity: darrinbaines.org

Verification date: 25 Nov 2025

Content version: 1.0.0

Canonical Identity

Term code
HE-ES-EA-012

Stable URI · Machine-readable · Resolvable · CC BY 4.0