Generalized Controlled Direct Effect for Multiple Mediation Analysis with Multiple Mediators
碩士 === 國立交通大學 === 統計學研究所 === 107 === Causal mediation analysis is a very prominent method to explore the mechanism of causality in recent years. It plays a paramount role in public policy, biomedical sciences, and economic issues. Recently, the theoretical framework for causal mediation analysis is...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | zh-TW |
| Published: |
2019
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/m9gs5g |
| Summary: | 碩士 === 國立交通大學 === 統計學研究所 === 107 === Causal mediation analysis is a very prominent method to explore the mechanism of causality in recent years. It plays a paramount role in public policy, biomedical sciences, and economic issues. Recently, the theoretical framework for causal mediation analysis is completed by the establishment of counterfactual framework. However, most of the current researches focus on indirect effect and direct effect for one mediator. In this article, we extend the definition of controlled direct effect from one mediator to multiple mediators in order to define the path-specific effect (gcPSE). All gcPSE can be identified by no unmeasured confounding assumptions and be estimated by the concept of difference method. Our method is also compared with existing methods for multiple mediation analysis.
|
|---|