Biostatistics Weekly Seminar

Mediation analysis seeks to describe how predictors affect outcomes through an intermediate variable. It can be used to test etiologic theory as well as to test the mechanism of action of an intervention. In the classic mediation analysis, the product of a and b path is the estimated mediated effect. However, in case of non-continuous (i.e., binary, ordinal categorical, count, nominal, survival time) mediator and/or outcome, this conventional product method is invalid. Counterfactually-defined mediation analysis provides an alternative way to examine the mediation effect with a non-continuous mediator and/or outcome, and the results are more interpretable. In this talk, I will review the concept and the calculations of counterfactually-defined mediation. An example analysis of using R and Mplus will be presented.

2525 WEA,10th floor, Room 1040
19 April 2019
1:30pm