Abstract:
|
Standard causal mediation analysis seeks to decompose a treatment or exposure effect into a direct effect and an indirect effect through a single mediator. Generalized causal mediation and path analysis (GCMPA), implemented recently in the GMediation R package, extends this methodology by accommodating multiple mediators at each of two causally-ordered ‘stages’ with mediators and the final outcome following generalized linear models. Approaches to sensitivity analysis, important for evaluating the impact of departures from sequential ignorability (or the no unmeasured confounders assumption), are currently available only for the single mediation situation. We extend a sensitivity analysis approach that allows outcomes following a generalized linear model to the GCMPA setting, thus, evaluating the effect of departures from sequential ignorability on path-specific effects. The method is illustrated using data from a study of causal pathways between socioeconomic status and childhood dental caries. Implementation of the new methods in the GMediation package is also discussed.
|