So my correlational results show that X is positively related to M (.70), and X and M are both negatively related to Y (-.52 and -.28).

All three correlations are as expected, and when I run the 'X -> M -> Y' mediation model, the indirect effect comes out to be significant. However, I see that the regression coefficients for the path b (association between M and Y controlling for X) are positive (.15).

I read that this is called inconsistent mediation, and my main question is if I can still consider this indirect effect as significant (and ok to report).

Also, I wonder if this is related to a ceiling effect (because the variable M IS highly negatively skewed towards the max).. or the association between M and Y being just too weak compared to X and Y...


This is also called "suppression," indicating that failing to account for one variable suppresses the effect of another variable on the outcome. What you have here are two opposing pathways from X: one from X through M to Y (the mediated pathway), for which higher values of X yield higher values of Y, and another from X to Y directly (or, more likely, through some other unmeasured mediator), for which higher values of X yield lower values of Y. This is fairly common but still very interesting. It's definitely still mediation, and definitely worth reporting. In fact, it's a much more interesting story than the traditional story of mediation. You have uncovered two opposing pathways from your predictor to your outcome.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.