|
Abstract:
|
In epidemiologic and genomic analyses, the mediation relationship among an extrinsic exposure (X), an intermediate variable or a set of intermediate variables (M), and an outcome (Y) is of rising interest. The sampling scheme of (X,M,Y) may be cross-sectional, or retrospective case-control sampling. While X is often considered as an instrumental variable that precedes (M,Y), the temporal (causal) order of M and X cannot be determined. For example, aberrant DNA methylation can be the cause or the consequence of cancer development. In this talk, I will discuss challenges of existing tests for determining the direction of causal mediation effect and propose potential remedies.
|
