Diffusion processes and event history analysis

Abstract

Several authors (e.g., Bruderl, Diekmann, Yamaguchi) derive hazard rate models of event history analysis from social diffusion processes. This paper also focuses on the integration of diffusion research and survival analysis. After a discussion of Diekmann's flexible diffusion model, we present an alternative approach which clarifies theoretical differences between popular rate models (e.g., the exponential model, log-logistic model, sickle model). Specifically, this approach provides a new rationale for the generalized log-logistic model in the sense of a flexible infection process. In cases with bell-shaped duration dependence, it thus allows a test for social contagion as a result of random contacts between actual and potential adopters. An application to divorce data serves as an illustration.