Limits...
Mediation analysis with intermediate confounding: structural equation modeling viewed through the causal inference lens.

De Stavola BL, Daniel RM, Ploubidis GB, Micali N - Am. J. Epidemiol. (2014)

Bottom Line: By giving model-free definitions of direct and indirect effects and clear assumptions for their identification, the latter school has formalized notions intuitively developed in the former and has greatly increased the flexibility of the models involved.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.Such confounders are naturally dealt with (albeit with the caveats of informality and modeling inflexibility) in the SEM framework.

View Article: PubMed Central - PubMed

Show MeSH

Related in: MedlinePlus

Structural equation model for exposure X, mediator M, outcome Y, background confounder C, and intermediate confounder L (error terms omitted for simplicity).
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4383385&req=5

KWU239F2: Structural equation model for exposure X, mediator M, outcome Y, background confounder C, and intermediate confounder L (error terms omitted for simplicity).

Mentions: Unlike the above, the definitions of direct and indirect effects given in the SEM literature depend on the specification of a particular statistical model (49). In the setting of Figure 2 (with single C and L), the following model for continuous Y, M, and L could be specified:(4){L=γ0+γxX+γcC+ϵlM=α0+αxX+αlL+αcC+ϵmY=β0+βxX+βmM+βlL+βcC+ϵy,where X and C are exogenous variables (no equations are specified for them), Y, M, and L are endogenous variables, and , , and are mean-zero error terms, uncorrelated with each other and with the exogenous variables. This is a linear path model for the joint distribution of Y, M, and L (4, 52).Figure 2.


Mediation analysis with intermediate confounding: structural equation modeling viewed through the causal inference lens.

De Stavola BL, Daniel RM, Ploubidis GB, Micali N - Am. J. Epidemiol. (2014)

Structural equation model for exposure X, mediator M, outcome Y, background confounder C, and intermediate confounder L (error terms omitted for simplicity).
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4383385&req=5

KWU239F2: Structural equation model for exposure X, mediator M, outcome Y, background confounder C, and intermediate confounder L (error terms omitted for simplicity).
Mentions: Unlike the above, the definitions of direct and indirect effects given in the SEM literature depend on the specification of a particular statistical model (49). In the setting of Figure 2 (with single C and L), the following model for continuous Y, M, and L could be specified:(4){L=γ0+γxX+γcC+ϵlM=α0+αxX+αlL+αcC+ϵmY=β0+βxX+βmM+βlL+βcC+ϵy,where X and C are exogenous variables (no equations are specified for them), Y, M, and L are endogenous variables, and , , and are mean-zero error terms, uncorrelated with each other and with the exogenous variables. This is a linear path model for the joint distribution of Y, M, and L (4, 52).Figure 2.

Bottom Line: By giving model-free definitions of direct and indirect effects and clear assumptions for their identification, the latter school has formalized notions intuitively developed in the former and has greatly increased the flexibility of the models involved.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.Such confounders are naturally dealt with (albeit with the caveats of informality and modeling inflexibility) in the SEM framework.

View Article: PubMed Central - PubMed

Show MeSH
Related in: MedlinePlus