Modelling Synergy using Manifest Categorical Variables

This paper discusses methods to model the concept of synergy at the level of manifest categorical variables. First, a classification of concepts of synergy is presented. A dditive and nonadditive concepts of synergy are distinguished. Most prominent among the nonadditive concepts is superadditive synergy. Examples are given from the natural sciences and the social sciences. M delling focuses on the relationship between the agents involved in a synergetic process. These relationships are expressed in form of contrasts, expressed in effect coding vectors in design matrices for nonstandard log-linear models. A method by Schuster is used to transform design matrices such that parameters reflect the proposed relationships. A n example reanalyses data presented by Bishop, Fienberg, and Holland (1975) that describe the development of thromboembolisms in women who differ in their patterns of contraceptive use and smoking. Alternative methods of analysis are com pared. Implications for developmental research are discussed.