TABLE OF CONTENTS The two previous examples differ in their conceptual approach from the notion of hierarchy of levels (aggregation effect and contextual effect), although they are both based on descriptive analysis methods. The explanatory approaches have produced methodological developments enabling a modeling of differentiations between individuals which takes into account the hierarchical structure of the data: multilevel statistical models [GOL 95], also more generally termed mixed models. These methods have been developed within microscopic analyses to move beyond one of the big "dilemmas" often experienced by human sciences modeling specialists: the choice of the most relevant level, microscopic or aggregated (see Chapter 6). They offer the possibility of analyzing the characteristics of a set of individuals keeping in mind the differentiations related to the environment or the context in which they evolve. There are many applications of these methods in the following fields:
in education sciences, where school results are modeled according to individual student characteristics and also by the characteristics of the class and the school [BRE 97, MOB 02] or the role of the shared family context in understanding differential parental treatment of children [JEN 03];
in epidemiology, where the probability of the prevalence of a pathology is a function of the inherent characteristics of the individuals and also of their environment [DUN 97, JON 95];
in voting behavior analysis, where the influence of context has been quite well demonstrated [JON 97];
in migration analysis, where the probabilities for migrating from one region to...
