Our method of constructing groups was, perhaps, not ideal; one might use latent profile models for a more elegant approach. However, we believe that the construction of groups from such models would have resulted in a somewhat different question than our method of forming groups. For example, both Booth-LaForce and Oxford (2008) and Eggum et al. (in press) found no high stable group of children for adult-reported withdrawal but did identify a group of children who repeatedly scored approximately 1 standard deviation or more above the mean on withdrawal. Booth-LaForce and Oxford (2008) noted that those children included some from both the increasing and the declining classes who remained relatively high in withdrawal despite increasing or decreasing somewhat in level. Both sets of investigators found that the high stable group was associated with dispositional, demographic, and/or family risk factors. Thus, identification of children high in problem behaviors over time (even if declining or increasing somewhat) can provide a somewhat different perspective on the development of adjustment problems than identifying children’s trajectories, especially when a high stable group is not identified using other methods.
