Effect decomposition in the presence of an exposure-induced mediator-outcome confounder.Epidemiology, 25, 300-6.
The potential reasons, that is, internal stresses, grain size effect, decomposition, ordering/disordering, are discussed.
Though requiring more assumptions, a major benefit of the natural effects is that effect decomposition works in the presence of interaction: you can obtain both an estimate of the direct and indirect effect.
First, these methods allow for effect decomposition in the presence of X-M interaction by defining direct and indirect effects (controlled or natural) from a potential outcomes (PO) framework and developing estimations of these quantities that are not model specific.
First, effect decomposition – the fact that the direct and indirect effects sum to the total – using the product or difference method only works in the special case where linear regression is used for the mediator and outcome models and when there is no exposure-mediator interaction.
Kohler, Karlson, and Holm propose a method (hereafter, "KHB method") for effect decomposition in a non-linear framework (Karlson & Holm, 2011).
However, through its predominant focus on nonparametric identification, the causal inference approach to effect decomposition via natural effects is limited to settings that exclude intermediate confounders.
Second, we implemented a hierarchical effect decomposition strategy to examine by which variables the effect of the intervention had been mediated.
Table 3 shows the results of the hierarchical effect decomposition, presenting crude and adjusted PRs (comparing infection after versus before the intervention).
Our estimate of joint effect decomposition is based on the hypothesized linear associations between birth weight and type 2 diabetes as well as between unhealthy lifestyle and type 2 diabetes, which may not be generalizable in other study populations.
