Ordinary least squares as a method of measurement

Abstract

The "classical" method of ordinary least squares which estimate two parameters of a linear (e.g. the demand) function is structuralistically reconstructed and is discussed from the point of theory of theories. The question is whether this statistical method can be subsumed under the notion of methods of measurement. The example used here is of course paradigmatic. The question in general is, whether a statistical procedure can be used in the right circumstances as a measurement procedure. We find arguments for the latter option.