CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: Relative-risk scale effects in mediation analysis Authors:  Monia Lupparelli - University of Florence (Italy) [presenting]
Alessandra Mattei - University of Florence (Italy)
Abstract: Causal mediation analysis studies the causal pathways of a treatment on an outcome by investigating the role of intermediate variables, named mediators, in the treatment-outcome relationship. When the effect of the treatment on the response might be channelled by mediators, the interest is on disentangling indirect effects, that are through the mediators of interest, and direct effects that are through other pathways other than the mediators. Under a sequential ignorability assumption, we propose a regression-based approach for binary outcomes and multiple mediators so that the total causal relative risk can be decomposed into the natural direct and indirect relative risk by combining model parameters.