Validity

Definition

Validity is the extent to which a research instrument, study design, or conclusion accurately represents the concept it is intended to measure or the causal relationship it claims to establish. It is not a single property but a family of concepts, including internal validity, external validity, construct validity, and content validity.

Why It Matters

A study can be statistically flawless yet invalid if it answers the wrong question or measures the wrong thing. Validity is therefore the ultimate criterion of research quality. Funding bodies, ethics committees, and journal reviewers evaluate validity before statistical significance. Without it, findings cannot be trusted, generalised, or translated into practice.

Example

A randomised controlled trial reports that a new teaching method improves exam scores. The study has high internal validity (randomisation controlled for confounders) and high construct validity (the exam accurately measures learning). However, it was conducted in a single elite private school. The external validity is low: the results may not generalise to state schools, different age groups, or other subjects. Policy-makers should therefore demand replication in more diverse settings before scaling the intervention.

Related Terms

• Construct Validity
• Content Validity
• Reliability
• Internal Consistency
• Cronbach's Alpha

Software Notes

• SPSS: Validity is assessed through a combination of procedures: correlation matrices (Analyze > Correlate > Bivariate), factor analysis (Analyze > Dimension Reduction > Factor), and SEM (AMOS). No single SPSS command computes "validity" as a single statistic.
• R: psych::fa() for construct validation via factor analysis. lavaan::cfa() and lavaan::sem() for confirmatory validation. psych::cor.ci() for convergent validity correlations.
• Stata: pwcorr for bivariate correlations supporting convergent validity. factor and sem for factor and structural validation. alpha for internal consistency, which is a prerequisite for validity.