ABSTRACT
Through the work of H. M. Blalock and Otis Dudley Duncan, sociologists were introduced to the practice of building recursive models of interdependencies among several variables and to the use of path analysis to provide a flow graph representation of such models and to interpret their estimation. This chapter identifies the term sensitivity analysis to a class of procedures for exploring the extent to which numerical estimates of parameters from an arbitrarily identified model depend upon the identifying assumptions. It shows that some possible avenues of research that could be utilized to make these procedures statistically rigorous. The chapter describes four methods for analyzing the sensitivity of parameter estimates in arbitrarily identified simultaneous-equation models to the identifying assumptions of those models. Hauser's model provides a good example of a nonrecursive simultaneous-equation model which is identified by arbitrary assumptions.
