Fig. 3.

Directed acyclic graph depicting the relationship among an exposure (X), a mediator (M) or multiple mediators (M = (M₁, …, M_m)), and an outcome (Y) in the classical mediation analysis. (A) Relationship between the exposure and the outcome, without considering the mediator. Here, c is the total exposure effect on outcome. (B) Relationship between the exposure, the mediator, and the outcome. Here, c′ is the direct effect of exposure on outcome, a is the exposure effect on the mediator, and b is the mediator effect on the outcome. The product of a and b (i.e., ab) represents the indirect/mediation effect. (C) Relationship between the exposure and the outcome with multiple mediators M = (M₁, …, M_m). Here, m is the total number of potential mediators; a = (a₁, …, a_m) is a vector of exposure effects on the mediators; and b = (b₁, …, b_m) is a vector of mediator effects on the outcome. The product of a and b (i.e., a^Tb) represents the indirect/mediation effects.