University of Pennsylvania
In mathematics, an argument is valid if its conclusion is logically entailed by its premises. In the social sciences, empirical research proceeds analogously from data to the scientific theories or answers research questions. However, social researchers tackle many more uncertainties than mathematicians do. For once, the social sciences are inductive and theorize statistical phenomena, hence social theories need to embrace probabilistic notions of validity. Moreover, if data are of questionable validity, their methodologically conclusive analysis is not likely to yield valid conclusions.
Validity should not be confused with → Reliability. Data are reliable to the extent to which the process of generating them is replicable. Replicability says nothing about what data are about. By contrast, data are valid to the extent to which they accurately represent the phenomena of analytical interest. A theory is valid to the extent to which it is corroborated by independently obtained evidence. In → Measurement Theory, a test is valid to the extent it measures what it claims it measure.
Four major kinds of validity can be distinguished. Logical validity concerns the conclusiveness of propositions derived from known premises. Logic has little to do with what resides outside its discourse but informs scientific argumentation and writing. Face validity is obvious or common truth. Researchers invoke face validity when they accept data, ...