Chapter 3: Longitudinal Modeling of Population Heterogeneity: Methodological Challenges to the Analysis of Empirically Derived Criminal Trajectory Profiles¹

The question of how to model population heterogeneity in longitudinal data has been repeatedly discussed in criminological journals, especially in the context of modeling “criminal careers” as the variable of interest (e.g., Elliot, 1985; Farrington & West, 1993; Nieuwbeerta & Blokland, 2003; Tracy, Wolfgang, & Figlio, 1990). It has been recognized that there is variation in criminal trajectories and any analysis of the age-crime relationship requires a decision on how to model this variation. In Chapter 1 in this volume, Muthén briefly touched on different modeling approaches to capture observed heterogeneity around trajectory profiles. Here we will revisit and extend the examples mentioned. To do so we will use two well-known criminological data sets, the Philadelphia Cohort Study (Tracy et al., 1990) and the Cambridge data (Farrington & West, 1990). Both datasets have been used in the past to describe criminal activity as a function of age.