By using random effects and latent class models, it should be possible to examine how suicidal and substance use behaviors interrelate with biological risk factors (e.g. 5-HT dysregulation), behavior (e.g. impulsivity and impulsive aggression), and environmental factors (e.g. stressful life events). Latent class models have been suggested as alternative methods for identifying subtypes of patients or substance users based on individual response patterns (e.g. Graham et al. 1991; Uebersax, 1994). Latent growth curve modeling is a statistical method for understanding growth patterns over time. More recently, an approach has been developed that combines latent class analysis and latent growth modeling (Muthén, 2001). Under this model, latent trajectory classes are derived that represent subgroups with similar growth trajectories or developmental patterns. The combined approaches are advantageous, since they allow classes to be defined based on their patterns of responses over time (Muthén, 2001). For example, these techniques were used to identify different patterns of aggressive responses to the “Good Behavior Game” prevention program for reducing aggression in the Baltimore city schools (Muthén, 2001). Such methods could be applied to identify groups
