View article

When the purpose of empirical analysis is to arrive at some provisional understand-ing of an observed phenomenon, one is attempting the selection of a mathematical representation of the phenomenon, that is, a model. As discussed in the previous chapter, numerous pitfalls await those who do not explicitly recognize this goal and properly account for it. The following article shows how this goal can be formally recognized, through the specification of a set of criteria that an acceptable model must satisfy, and how to account for this goal, through the application of a prescribed set of specification tests–in particular m-tests. The setting is compatible with that of Chapter 5 and with White (1994), and the specification criteria and associated tests are, for the most part, those expounded in the foregoing chapters. The approach described next is just one possibility. Another is treated in the final chapter. For a collection of …