The output of the mediation package (single mediator) has given me an ACME that is negative and statistically significant, an ADE that is positive but not significant and a total effect that is negative but not significant. How do I interpret this result?
mediate(modelM, model Y, treat=X, mediator=M, boot=TRUE, sims=500)
The effects through the mediator and not through the mediator essentially cancel each other out. You have power to detect the mediated effect but not the direct effect. The total effect may indeed be 0, or it may be some value other than 0. All you can claim is that the effect through the mediator is negative. You can't say anything about any other effects if they are not significant.
