How to Design a Panel Study for Investigating Bivariate Dynamics Using Continuous-Time Modeling

Designing effective longitudinal panel studies to capture bivariate dynamics requires making informed decisions about how many measurements to collect and when to collect them. This study examines how design choices, concretely sample size, number of repeated measures, and various aspects of the temporal sampling schedule, influence the recovery of bivariate dynamic parameters in panel data analyzed using a Continuous-Time Random Intercept Cross-Lagged Panel Model (CT-RI-CLPM). Through a Monte Carlo simulation, we tested a range of generating population parameters. We proposed a novel approach for planning sampling schedules: the Short-and-Long Lags design (SHALL) which deliberately mixes closely spaced and widely spaced intervals that continuous time models can accommodate. SHALL schedules are simple to understand and implement, and consistently ranked among the best sampling schedules in many scenarios. Our findings suggest that the number of repeated measurements and sample size significantly impact parameter recovery, particularly for cross-lagged effects, which exhibited low estimation efficiency even in the most favorable conditions, leading to low statistical power. We provide actionable recommendations for researchers planning panel data studies, along with open-source R code to facilitate the design of SHALL sampling schedules. We also provide R code to implement a CT-RI-CLPM to new data sets. An empirical example illustrates the application of such a model to data from an empirical panel study relating self-esteem and depression. This work contributes evidence-based guidance for improving the design and analysis of longitudinal panel studies using continuous-time modeling.