Bayesian dynamic mediation analysis
Most existing methods for mediation analysis assume that mediation is a static, time-invariant process, which overlooks the inherently dynamic nature of many human psychological processes and behavioral activities. In this dissertation work, we consider mediation as a dynamic process that continuously changes over time. We propose Bayesian time-varying coefficient model to describe and estimate such dynamic mediation effects. By taking the nonparametric penalized spline approach, the proposed method is flexible and able to accommodate any shape of the relationship between time and mediated effects. By modeling mediation effects nonparametrically as continuous functions of time, the dynamic mediation model provides a valuable tool to help researchers obtain a more complete understanding of the dynamic nature of the mediation process underlying psychological and behavioral phenomena. Heterogeneity and multi-level structures exist in almost every study populations. We extend the dynamic mediation model to propose 1) a latent class dynamic mediation model which is able to detect heterogeneity by the differences in characteristics of mediation mechanisms among unobserved subclassese within a study population, and 2) a hierarchical dynamic mediation model which can accommodate data with at least 3 level hierarchical structures using random effects model to estimate dynamic subject-specific mediation effects and cluster-specific mediation effects. Simulation studies show that the proposed methods work well and faithfully reflect the true nature of mediation process.
Huang, Jing, "Bayesian dynamic mediation analysis" (2016). Texas Medical Center Dissertations (via ProQuest). AAI10126223.