Generalized Causal Mediation and Path Analysis and Its R Package “gmediation”

• Activity Number: 531 - SPEED: Statistical Computing: Methods, Implementation, and Application, Part 2
• Type: Contributed
• Date/Time: Wednesday, July 31, 2019 : 11:35 AM to 12:20 PM
• Sponsor: Section on Statistical Computing
• Author(s): Jang Ik Cho* and Jeffrey M Albert
• Companies: Eli Lilly and Company and Case Western Reserve University
• Keywords: mediation; path analysis; gmediation; R; multiple mediators

Abstract

Mediation analysis provides a way to understand the underlying mechanism through which an exposure affects an outcome. Hence, mediation analysis is widely used in finding causal relationships in many fields including health sciences. Mediation analysis methods have been rapidly developed in the last few decades yet most of the methods, and the few relevant R packages, focus on a single `stage' of (that, is non-sequential) mediation. In reality, studies often involve multiple mediators occurring in multiple stages, providing many, and longer, possible paths. To address this more general situation, we developed a generalized mediation formula method (Albert, Cho, Liu, and Nelson 2018) implemented in an R package called gmediation. This new package can handle an arbitrary number of mediators of multiple possible types (binary, count or continuous) occurring at two stages. This presentation/poster will give an overview of our method and the usage of the gmediation package, illustrated with a dental data example.