A Tutorial on Bollen and Brand's Approach to Modeling Dynamics While Attending to Dynamic Panel Bias

Recent studies demonstrate that when researchers are interested in dynamics they are better off using a statistical model described in Bollen and Brand (2010) rather than the often employed random-coefficient or multi-level model (Moral-Benito et al., 2019; Xu et al., 2020). Their Monte Carlo studies, however, were methodologically advanced papers. Here, we present a beginner, hands-on tutorial describing the technique. We provide code in snippet form that any researcher can apply to his or her longitudinal data, introduce fundamentals of dynamic modeling, and generalize the basic model in Bollen and Brand (2010) to situations that cover broader inferences than those discussed in the simulation articles. Translational Abstract Longitudinal data are now common, and there are many statistical models that researchers can use to assess the patterns present in panel data to make inferences about dynamics. This article offers a tutorial on dynamic modeling using an approach described by Bollen and Brand (2010). Our target audience includes both graduate students and career academics who are interested in dynamic modeling but have not yet explored the topic or run their own analyses. Our goal is to gently guide readers through dynamics concepts and code, leaving them ready to start modeling on their own.

Automated summary

Recent studies suggest that using the statistical model described by Bollen and Brand (2010) for dynamics is more effective than commonly employed random-coefficient or multi-level models. While previous Monte Carlo studies were methodologically advanced, this article provides a beginner-friendly tutorial on dynamic modeling, introducing code and fundamentals to guide readers through the process.