I have a significant interaction $A\times B$ in my regular mixed ANOVA, but when I add a covariate, the effect $A\times B$ is no longer significant, but the interaction effect with the covariate $A\times B\times \mathrm{covariate}$ is. Why is this the case?

  • 1
    $\begingroup$ This is because the two-way interaction $A\times B$ is different across the values of $\mathrm{covariate}$ $\endgroup$ – Robert Long Mar 6 at 17:00
  • $\begingroup$ Thank you. Would this be considered to be some kind of mediation effect? $\endgroup$ – user48535 Mar 6 at 17:02
  • $\begingroup$ I hope not. Is $\mathrm{covariate}$ on the causal pathway from A/B to the outcome ? If so then it is a mediator and should not be included in the model if your interest is in inference. $\endgroup$ – Robert Long Mar 6 at 17:05
  • $\begingroup$ No, that wouldn't make any sense. Does this kind of effect go by any specific name I could mention in the study? $\endgroup$ – user48535 Mar 6 at 17:14
  • $\begingroup$ Not really, this is a fairly common occurrence with interactions. It's basically the same as a significant 2-way interaction but insignificant main effect(s). It is important to understand that the meaning of the main effect for a variable (or two-way interaction in your case) is not the same in the presence of a 2-way interaction involving that variable(or 3-way interaction in your case). There are lots of questions and answers about this here on CV, such as this $\endgroup$ – Robert Long Mar 6 at 17:32

Your Answer

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

Browse other questions tagged or ask your own question.