Rechercher

Sur ce site


le mercredi 26 septembre à 12:15

Séminaire Biostatistique - Pr B. RACHET

par Vincent Daubin - 26 septembre 2018

erence and mediation analysis : an overview of the principles and assumptions

Randomised clinical trials (RCTs) are the gold standard to make causal inference on associations between an exposure/treatment variable and an outcome. However, RCTs are not (ethically or practically) feasible in many situations. Using the counterfactuals or the potential outcomes, important conceptual development has been conducted in the last decades around causal inference derived from observational data.
The lecture will first describe the principle of causal association and the key assumptions behind such an association, then their translation into observational data. Some of the main approaches used to estimate causal association will be briefly presented. Furthermore, one may be interested in decompose the causal association into different pathways though mediators. We will see how to identify various direct and indirect (‘mediated’) effects, as well as the additional assumptions required to estimate such effects. Some approaches used to estimate them will be presented alongside real-life illustrations.

L’inscription est gratuite mais obligatoire et doit se faire au plus tard le jeudi 20 septembre 2018 par courriel à paola.damaso@chu-lyon.fr