TIGP (BIO)—From Linear Structural Equation Modeling to Generalized Multiple Mediation Formula—Keywords: causal inference, mediation analysis, multiple mediation, SEM, counterfactual, interventional approach.
- 講者林聖軒 教授 (國立陽明交通大學 統計學研究所)
邀請人：TIGP (BIO) - 時間2022-05-12 (Thu.) 14:00 ~ 16:00
- 地點僅提供視訊
線上串流
摘要
Causal mediation analysis is advantageous for mechanism investigation. In settings with multiple causally ordered mediators, path-specific effects (PSEs) have been introduced to specify the effects of certain combinations of mediators. However, most PSEs are unidentifiable. Interventional analogue of PSE (iPSE) is adapted to address the non-identifiability problem. Moreover, previous studies only focused on cases with two or three mediators due to the complexity of the mediation formula in large number of mediators. In this study, we provide a generalized definition of traditional PSEs and iPSEs with a recursive formula, along with the required assumptions for nonparametric identification. This work has three major contributions: First, we developed a general approach (that includes notation, definitions, and estimation methods) for causal mediation analysis with an arbitrary number of multiple ordered mediators and with time-varying confounders. Second, we demonstrate identified formula of iPSE is a general form of previous mediation analysis. It is reduced to linear structural equation model under linear or log-linear model, to causal mediation formula when only one mediator. Third, a flexible algorithm built based on g-computation algorithm is proposed along with a user-friendly software online. This approach is applied to a Taiwanese cohort study for exploring the mechanism by which hepatitis C virus infection affects mortality through hepatitis B virus infection, abnormal liver function, and hepatocellular carcinoma. All methods and software proposed in this study contribute to comprehensively decompose a causal effect confirmed by data science and help disentangling causal mechanisms when multiple ordered mediators exist, which make the natural pathways complicated.
BIO
I am an associate professor in the Institute of Statistics and Institute of Data Science and Engineering at National Yang Ming Chiao Tung University. I received the degree of Medical Doctor (MD) from National Taiwan University in 2010, Master of Science (ScM) from Department of Biostatistics and Doctor of Science (ScD) from Department of Epidemiology at Harvard University in 2016, and spent one year at the Department of Biostatistics at Columbia University as a postdoctoral Research Scientist between 2016 and 2017.
I am the recipient of the 9th Young Scholar Creation Award (from Foundation For The Advancement of Outstanding Scholarship), of the 2018 MOST Young Fellowship (Grant of Columbus Program) from Ministry of Science and Technology in Taiwan, of the 2016 Reuel Stallones Student Prize Paper Award from Society of Epidemiology Research (SER), and of the 2019 Young Scholar Research Award from the College of Science at National Chiao Tung University.
I focus my research in the area of Epidemiologic methods, employing causal inference theory to formalize concepts of mediation and pathway analysis and developing statistical and computational methods for effect decompositions. I am currently working on the methods for causal mediation analysis with time-varying and multiple mediators in complicated systems, extending mediation methods to longitudinal settings, along with corresponding software development. On the applied side, I am actively applying the new methods to investigate the causal mechanism for diseases including Alzheimer's disease, cardiovascular disease, and the relationship between religion and health. I am also collaborating with clinicians in a neuropsychiatric disorder such as sleep deprivation and behavioral addiction. I have published 52 journal papers, 28 as first or (co-) corresponding author (till April 2022).