Méthodes d'analyse du changement fondées sur les trajectoires de développement individuelle : Modèles de régression mixtes paramétriques et non paramétriques[1]

Abstract

Longitudinal Methods Based on Individual Development Trajectories - Parametric and Non Parametric Mixed Models: Generalized linear mixed models encompass a variety of modern longitudinal analytic approaches based on individual developmental trajectories. These models overcome many important problems inherent to other traditional analysis of longitudinal data. They all rely on two basic levels: the lower one express, through a set of parameters, the individual pattem of change over time (within-individual change), whereas the upper level captures the variations between these parameters describing individual trajectories (between-individual differences in change). However, other characteristics distinguish différent sorts of mixed models, such as their assumptions concerning the distribution of the trajectories within the population. This introductory article presents the basic linear mixed model assuming a normal distribution of the unobserved heterogeneity, and the nonparametric mixture model that relies on a discrete approximation of the unobserved heterogeneity. Before comparing these two models, the first section of the article gives a general description of the notion of individual developmental trajectories.