Causal Mediation Analysis Under Partial Compliance in Single-Site and Multisite Randomized Trials

The purpose of this project is to develop analytic strategies that incorporate instrumental variables for identifying and estimating the average treatment effect (ATE), the moderated ATE, the mediated ATE, and the moderated mediation effects in a single-site randomized trial where the moderator of theoretical interest is individual predisposition for compliance. The research team will also develop approaches to causal moderation and mediation in cluster randomized trials to take into account site-level compliance and its potential effect on treatment effect heterogeneity in the ATE and in mediator effects. Extending the new set of approaches for cluster randomized trials, this study will compare a method-of-moment estimation approach with Bayesian random-effects models.