Session: Applied Longitudinal Models for Concurrent Developmental Processes in Prevention Research (Society for Prevention Research 21st Annual Meeting)

Longitudinal modeling is well ensconced in prevention research, with several classes of statistical models available for the most common longitudinal design: the discrete-time (panel) study of single or multiple cohorts. Most research, however, has focused on single processes, i.e., one outcome as it unfolds over time. To the extent that multiple longitudinal processes are examined, they are often modeled in separate analyses, precluding empirical evaluation of the relations between different outcomes over time.  When processes are modeled together, one process is typically “privileged” over the other, such that the effects linking the two are not reciprocal. This symposium focuses on illustrating the use of models drawn from a broader class of concurrent- or dual-process longitudinal models in which two longitudinal outcomes co-occur and the model accommodates effects of either process on the other.

The first paper in this symposium presents a unifying framework for the diffuse methodological work that has already been done for the simultaneous modeling of joint longitudinal processes.  A model taxonomy is proposed that highlights both the similarities and differences across existing joint longitudinal models and identifies dual-process settings for which appropriate analytic models are either under-developed or under-applied.  The first paper concludes with an emphasis on the novel features of the dual-process models to be demonstrated in the remaining papers. 

The second paper illustrates the application of dual-process ordinal growth models to investigate the reciprocal relationships between substance use and high-risk sexual behavior across adolescence and emerging adulthood.  Comparisons are made between dual processes involving high-risk sexual behaviors and alcohol, tobacco, and marijuana use, respectively.  The third paper utilizes a dual-process discrete-time survival analysis to characterize the reciprocal hazard risk of initiation of substance use (alcohol and tobacco use) and high school dropout.  This modeling approach permits simultaneously testing the mediation of the effect of parental monitoring on the hazard risk of substance use onset by high school dropout and the mediation of the effect of parental monitoring on the hazard risk of high school dropout by substance use initiation. The symposium concludes with a discussion of other potential uses of dual-process models in prevention research, highlighting future opportunities for both substantive and statistical advancements related to joint longitudinal processes.