criterionrelated

Criterion-related, or criterion-related validity, is a term used in measurement theory to describe the degree to which a measurement instrument relates to an external criterion considered to reflect the intended outcome. The adjective is often written as criterion-related, and the noun form is criterion validity.

It is demonstrated by examining the statistical association between the instrument scores and the external criterion. A strong relationship provides evidence that the instrument can predict or reflect the criterion in a meaningful way. There are two common forms: predictive validity, where the criterion is measured after the instrument is administered to test future outcomes (for example, a job aptitude test predicting later job performance); and concurrent validity, where the criterion is assessed at roughly the same time as the instrument (for example, correlating a new test with an established measure collected in the same session).

Assessment typically uses correlation coefficients (such as Pearson r or Spearman rho), regression analyses, or, when the criterion is binary, ROC analysis to evaluate discrimination. Key considerations include ensuring the criterion is relevant, reliably measured, and independent from the instrument being evaluated, as well as sample size and potential range restriction.

Criterion-related validity is a component of construct validity and is distinct from content validity, which concerns alignment with the domain of the construct, and from convergent and discriminant validity, which examine relationships with related and unrelated constructs, respectively. Limitations include that poor or changing criteria can undercut validity, and high correlation does not guarantee generalizability across contexts or populations.