ABSTRACT
The study of dynamic processes calls for the formulation of models that explicitly define the nature of the dependency on time. Methods for event history analysis are designed to estimate the parameters of such processes including the effects of time and of characteristics of the individuals. These methods are more suitable than conventional regression techniques inasmuch as they do not depend on rather doubtful assumptions about the equilibrium of the process. The utility of these methods, however, depends on the solution of four central problems: treatment of censored observations, measurement of time (discrete/continuous), estimation of effects of time-varying covariates, and heterogeneity of observations. In this paper we describe the nature of each of these problems and review alternative solutions.
