Ordinal Nomological Validity
Ordinal Nomological Validity Assessment · Also known as: nomological validity for ordinal data, ordinal nomological network, construct network validity (ordinal), ordinal criterion-related validity
Ordinal nomological validity examines whether a construct measured with ordinal items (e.g., Likert-type scales) behaves in theoretically predicted ways within a nomological network — a web of expected relationships with other constructs and criteria — using methods suited to ordinal data rather than assuming continuous measurement.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use ordinal nomological validity when validating an ordinal scale (Likert, rating, frequency categories) and theory specifies relationships with external constructs. It is particularly important when the construct is novel or adapted from another language or culture. Do NOT use this approach as a substitute for confirmatory factor analysis — it supplements, not replaces, internal structure evidence. Do not rely on Pearson correlations alone for ordinal data when response categories are few (e.g., 1–4 scale) or skewed, as attenuation bias will underestimate relationships and may mislead validity conclusions. This approach also requires a well-developed nomological network; if no prior theory specifies expected relationships, convergent and discriminant validity procedures are more appropriate starting points.
Strengths & limitations
- Directly tests theoretical predictions, linking validation to substantive theory rather than purely statistical criteria.
- Polychoric and WLSMV-based approaches correct for attenuation from ordinal discretisation, yielding more accurate estimates than Pearson-based methods.
- Embeds validation within a broader construct network, demonstrating not just reliability but theoretical coherence.
- Ordinal SEM methods provide model-fit indices that allow formal evaluation of the entire pattern of relationships simultaneously.
- Generalises well across diverse ordinal formats including Likert, Thurstone, and forced-choice response scales.
- Requires a well-articulated nomological network before data collection; post-hoc theorising weakens evidential value.
- Polychoric estimation can fail or produce improper solutions with small samples or sparse ordinal categories.
- DWLS/WLSMV estimators require relatively large samples (often n > 200) for stable model-fit statistics.
- The assessment is inherently theory-dependent: a poorly specified nomological network yields uninterpretable results.
- Results can be confused with convergent validity, but nomological validity specifically tests theoretically derived patterns, not merely correlational overlap.
Frequently asked
How is ordinal nomological validity different from ordinary nomological validity?
The conceptual goal is identical — verifying that a construct behaves as theory predicts within a network of related constructs. The difference is methodological: ordinal nomological validity uses polychoric correlations, rank-based coefficients, and ordinal SEM estimators (DWLS, WLSMV) instead of Pearson correlations and ML-based SEM, correcting for the attenuation and distributional distortions that arise when ordinal responses are treated as continuous interval scores.
How many criteria should the nomological network include?
A network with at least two convergent (positively related) and one discriminant (weakly or negatively related) construct is a minimum for a credible assessment. More criteria strengthen the argument, but each relationship must be grounded in theory specified before data collection, not selected after examining results.
Can I use Spearman correlations instead of polychoric correlations?
Spearman's rho is a non-parametric option and is appropriate when the relationship may be monotone but not linear. However, when the constructs are assumed to reflect underlying normal latent variables — which is the typical psychometric assumption — polychoric correlations more accurately recover the latent-level association. Spearman tends to still underestimate compared to polychoric for coarsely categorised responses.
What sample size is needed for ordinal SEM nomological validity?
DWLS and WLSMV estimators are more robust to non-normality than ML but still require adequate samples. Common guidance suggests at least n = 200 for simple models; complex nomological networks with many constructs may require substantially more. Simulation studies (e.g., Finney & DiStefano, 2006) suggest WLSMV maintains acceptable performance at smaller samples compared to DWLS for highly non-normal ordinal data.
Should nomological validity be assessed before or after establishing measurement invariance?
If nomological relationships are examined across subgroups (e.g., gender, cultural groups), measurement invariance must be established first. Without at minimum metric (loadings) invariance, observed differences in nomological relationships may reflect measurement artefacts rather than true construct-level differences.
Sources
- Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological Bulletin, 52(4), 281–302. DOI: 10.1037/h0040957 ↗
- Borsboom, D., Mellenbergh, G. J., & van Heerden, J. (2004). The concept of validity. Psychological Review, 111(4), 1061–1071. DOI: 10.1037/0033-295X.111.4.1061 ↗
How to cite this page
ScholarGate. (2026, June 3). Ordinal Nomological Validity Assessment. ScholarGate. https://scholargate.app/en/psychometrics/ordinal-nomological-validity
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Confirmatory factor analysisPsychometrics↔ compare
- Nomological ValidityPsychometrics↔ compare
- Ordinal Convergent ValidityPsychometrics↔ compare
- Ordinal Discriminant ValidityPsychometrics↔ compare
- Ordinal Measurement InvariancePsychometrics↔ compare