The Interaction of Measurement and Structure in Simultaneous Equation Models with Unobservable Variables

The authors report the results of some theoretical investigations of interactions that occur in the simultaneous estimation of parameters. Under the assumptions of a perfectly specified model for a set of non-overlapping indicators and no sampling fluctuations, an overall consistency criterion is developed showing that though measurement and structure can be evaluated independently of each other, in general, they should not be. The proposed consistency criterion clearly shows that the internal and external consistency criteria prescribed by Anderson and Gerbing are not sufficient; that is, there are correlation matrices for which the criteria of internal and external consistency both can be satisfied, yet the matrix as a whole is inconsistent. Implications of the interaction between measurement and structure are also discussed.