A developmental trajectory describes the course of a behavior over age or time. This project builds upon prior NSF-funded research (SES-9911370 & SES-9511412) that developed a group-based approach for identifying distinctive groups of individual trajectories within the population of interest. As part of that research, "canned" SAS-based software (Proc Traj), distributed free of charge, was developed for estimation of the trajectory models. This project will address three major issues. First, all longitudinal studies, particularly of people, suffer from the problem of missing assessments because subjects drop out of the study for reasons such as death or because they can not be located or are no longer willing to participate. Because subject drop-out is not random, one important objective of the research is to extend the basic model to account for differential drop-out rates by trajectory group. This enhancement will be incorporated into the Proc Traj software. Second, prior work links group-based trajectory modeling with work on propensity scores and matching. The aim of this linkage is to make (more confident) causal inferences about the effect of major life events or therapeutic interventions on trajectories of behavior. The posterior probabilities of group membership, a key product of group-based trajectory modeling, play a central role in linking these two lines of research. A second important objective of the research is to explore the contribution of the posterior probabilities in creating balance on measured covariates between treated and controls. Third, for reasons of tractability, trajectory models estimated with Proc Traj assume conditional independence across time periods within trajectory group. This research also will explore the impact of this assumption's failure on parameter estimates and their standard errors.
The methodological extensions will be used to investigate important substantive questions in developmental psychopathology and developmental criminology. In this regard, the work on causal inference has an important application domain in assessing the effectiveness of nonrandomly assigned medical treatments, including psychiatric counseling and drug treatment, and in analyzing the impact of events such as incarceration on subsequent criminality. Application of the extension to account for non-random drop-out may reveal important biases in findings that do not take this phenomenon into account. Also, wide dissemination of Proc Traj free of charge allows other researchers to more easily utilize the methodological advances of this research. It is important to note that many Ph.D. candidates and postdoctoral scholars use Proc Traj. Thus, this research contributes to the education of the next generation of researchers.