Moderator analyses within a trajectory framework

• In Event: Extensions to Group Based Trajectory Modeling

Abstract

Much of the research on turning points estimate the average relationship between life events and subsequent criminal behavior. While it is important to estimate the unbiased effect of potential turning points on criminal behavior, many of these studies overlook treatment effect heterogeneity and the extent to which there is variation from average relationship reported (Nguyen & Loughran, 2017). This may oversimplify the relationships in question. Trajectory models offer a way to describe heterogeneity in behavior as well as heterogeneity in the impact of life events. Using a dual trajectory model to link patterns of arrest before and after a prison term, this paper considers treatment effect heterogeneity in the extent to which prison serves as a turning point in the criminal career. Specifically, two distinct moderator analyses are employed to determine how inmate age, race/ethnicity, gender, prior incarceration history, and time-served condition transitions between pre- and post-prison arrest trajectories. The analyses performed include and extend beyond the traditional post-estimation multinomial logistic regressions to provide a different way to model treatment effect heterogeneity within a trajectory framework. Because there is potential error in the dual trajectory estimation, several ways to account for model specification and classification errors are also explored.

Author

• Nicole Frisch, Merrimack College