On deriving models in the social sciences

Five different kinds of derivation are examined. The first are measurement representations, the second are regression-type models, the third are non-Markov observable models with process assumptions. The fourth are discrete Markov models with unobservable variables playing a theoretical role, and finally, the fifth return to the classical derivation of differential equations so characteristic of the physical sciences. Also discussed are some general issues raised by these various methods of derivation. I touch at least briefly on questions of axiomatization, the use of theoretical or unobservable variables, and the relative importance of measurement procedures.