Target Mean Analysis — Direct Effect

Context

• The Target Mean Analysis used to calculate Direct Effects is closely related to the Target Mean Analysis — Total Effects.

  • Using the simulated x-values across the range of the variable’s variation domain, Target Mean Analysis infers the corresponding mean values of the Target Node, i.e., the y-values in the plot, while holding constant (or fixed) the distributions of the other variables in the network.
    • For instance, when computing the curve of the Target Node $\mathit{Pleasure\_(4)}$ as a function of $\mathit{Intensity}$, the marginal distributions of the other variables, $\mathit{Pleasure}$, $\mathit{Corresponds}$, $\mathit{Easy\ to\ wear}$, are maintained.
    • Fixing distributions does not apply if any of the nodes belong to the following classes:
      • Factor or
      • Non_Confounder

• In the following plot, the Direct Effect Curves of all Driver Nodes are displayed at the same time.

• However, please note that all of these curves are calculated individually.