Chen Liu received the Joint Statistical Meetings Student Paper Award from the Health Policy Statistics Section of the American Statistical Association for "Heterogeneous Causal Mediation Analysis Using Bayesian Additive Regression Trees,” a paper that introduces a novel approach revealing individual-level variations in mediation mechanisms underlying treatment effects on health outcomes, offering insights that could lead to more effective health interventions.
Traditionally, causal mediation analysis has focused on average effects for the entire population, often overlooking critical individual-level differences. Liu's research addresses this limitation by employing a Bayesian regression tree ensemble to flexibly model complicated, non-linear relationships and capture the interactions between treatments, mediators, and other factors that affect outcomes. Using hierarchical posterior sampling, Liu generates credible intervals for heterogeneous mediation effects, ensuring that individual variations are accurately inferred.
Through detailed simulations, Liu demonstrated the accuracy and effectiveness of this method in estimating the heterogeneous mediation effects. To show the real-world impact, Liu applied this method to Alzheimer's disease research, specifically examining how Alzheimer's pathology mediates the effect of the APOE gene on cognitive decline in later life. This shows how Liu's method can answer complex health questions and improve personalized medical care.
This August, Liu will present his research at the Joint Statistical Meetings of the ASA in Nashville, Tennessee and receive his award. He is advised by Jiebiao Wang and Xu Qin.
-Calvin Dziewulski