Using Cluster Analysis for the Identification of Multivariate Trajectories

Julien Morizot, Universite de Montreal
Marc LeBlanc, University of Montreal

Criminologists have used six strategies for studying quantitative and qualitative changes in the course of offending. They provided interesting information on the various developmental trajectories of offending. However, all of them suffer of some limitations. The most often used method to date was the semi parametric group-based modeling on a single variable. In this paper, we will present an alternative analysis to the common groups detection methods used in criminology: cluster analysis. Even if clustering techniques were subjected to criticisms, we will show that when used in an appropriate way, this analytical strategy could overcome the univariate nature of the methods used to date and allow researcher to analyze small special population samples.

(Return to Program Resources)

Updated 05/20/2006