The theory of a concept dictates which statistical measure of it is valid.

Evidence

The analysis in this section has attempted to adjudicate between conflicting arguments regarding the substitutability of trade dependence measures. If the tenor of bilateral relations is primarily determined by the less dependent—and, hence, less constrained—state and the more dependent state plays no compensatory role at all, then a weakest-link measure of interdependence should be most effective. But if a more dependent actor recognizes its position and take pains to moderate its behavior so as not to provoke a less dependent partner, then an appropriate measure must accommodate some substitutability. The evidence adduced here, while by no means definitive, has highlighted the efficacy of the geometric mean, thus suggesting at least some measure of substitutability. — Gary Goertz, Social Science Concepts, 974

Categorization

• DEEPTHINKING: A genuinely load-bearing methodological claim: measurement is not neutral arithmetic but an encoding of assumed causal substitutability — choose the mean and you have chosen a theory.
• XREF: Recalls Milgrom’s account of reputation institutions and the diamond bourse, where the choice of metric (reputation vs. formal contract) presupposed a theory of enforcement.