Time-lagged panel models in psychotherapy process and mechanisms of change research: Methodological challenges and advances

Abstract

In recent years, there has been increasing interest in utilizing time-lagged panel models to study mechanisms of change in psychotherapy. These models offer valuable insights into the dynamic relationships between variables over time and offer stronger causal inference capabilities than cross-sectional analyses. Therefore, they are well-suited for modeling the intricate relationships between mechanisms of change and outcomes in psychotherapy studies, which are typically beyond experimental control. However, their complexity, coupled with the fact that detailed explanations are often embedded in dense statistical or econometric texts, poses challenges. This paper provides a background on cross-lagged panel models and delves deeper into explaining the issues of 1) dynamic panel bias, 2) long-run effects, and 3) testing whether different treatments work by different mechanisms. Using data from a psychotherapy study on treatment of adolescent depression, I demonstrate how these issues manifest in real data. In conclusion, I recommend using structural equation modeling to circumvent dynamic panel bias, reporting long-run effects to reveal the long-term impact of sustained therapeutic work on mechanisms of change, and carefully considering whether mediation, moderation, or a combination of both, best describes differential effects of mechanisms between treatments.