PM in no interaction

[jhcreed]

Proportion mediated differs largely from the values from SAS when there is no interaction between exposure and mediator. See tests-no-interaction. However the PM from SAS doesnt necessarily make sense either. For example in Explanation In Causal Inference pg 47 it says "When the effects are used on the difference scale, the proportion mediated is then just defined as the ratio of the natural indirect effect to the total effect, that is, PM = NIE/TE". Using the continuous outcome/mediator example in tests-no-interaction the SAS macro returns a NIE = -0.001643, TE = 0.071753 and PM = 0.49434 but -0.001643/ 0.071753 = -0.022898 not 0.49434. So different rules when there is no iteration?

Please double check @tgerke ?