A Rejoinder on Models, Metaphors, and Fables

Julie Shaffer deserves our thanks for putting this set of essays together, and the discussants have obviously taken the project quite seriously. Keith Hope, in particular, gets a lot of credit not only for scholarship but also for good humor under fire. I will respond to the discussants; give a few examples (from epidemiology) where causal inferences have been drawn from observational data, but not by path-analytic techniques; and summarize my views. The focus of the rejoinder, as of the essay, is on the hinge connecting the path models to reality-the assumptions. The main assumption, that the equations are structural, is explained in my research fable; the statistical assumptions are outlined in the sections on regression and path models.