I have a mediation model wherein the total effect (x -> y) is non-significant, but the direct effect ox x->y is significant. There is no significant indirect effect (no mediation), and the IV- mediator relationship, and mediator-DV relationship (path a and b) are non significant. I don't know how to interpret or report this, since my original hypothesis (that X significantly relates to y, via correlations and comparing Y in groups scoring low and high on X, and incidentally a NS total effect in the mediation output on PROCESS) was not met, but when I test for mediation, suddenly there is a direct effect of X on Y, and there is no evidence of the mediator relating to X nor Y!

How can this be interpreted?



1 Answer 1


There are many considerations that are important. One is whether the results are suggesting partial or complete mediation. One early consideration to rule out is that significance should not be determined entirely on p-values meeting a particular threshold (0.05, 0.1, etc). Assuming you are not claiming non-significant for a p-value that is close but does not meet one of those thresholds (e.g. p = 0.07 or p = 0.15), let's move forward.

Per Baron & Kenny (1986), there are three sequential steps for testing mediation.

  1. Check if there is a significant relationship between X and Y (i.e. Path C).

  2. If yes to (1), examine whether X and the mediating variable are significantly related (i.e. Path A).

  3. If yes to (2), examine whether the mediating variable and Y are significantly related (i.e. Path C) but this time while controlling for X. If this coefficient is non-significant, then you have a case of full mediation. If it is significant, you have a case of partial mediation.

enter image description here

In your case, I understand that your models fail step 1. While the steps above suggest this is reason enough to call no mediation, increasing research argues that the total effect does not need to be significant. Please see this answer for an absolutely excellent overview.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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