I need to test the effect of X on Y through M. Based on knowledge, I know X does not have direct effect on Y. X has to act on Y through some mediator(s). In this case, do I still need to estimate the direct effect of X on Y (c') in my model? Is it ok if I only estimate a (effect of X on M) and b (effect of M on Y), and test if the product of ab is significant?


Most mediation tests require two models and you do not clarify which you are referring to. If you use the Baron/Kenny approach, you need two models: X->M and M->Y. If you do the Sobel test, you will fit a model with the marginal effect X->Y as well as the conditional (direct) effect X,M -> Y.

  • $\begingroup$ Thanks for your answer. My question is on the last model "conditional (direct) effect X, M ->Y". Based on my knowledge, if I know X has no way to act on Y directly, can I skip the last model? Is it OK is I only have two models X->M and M->Y? $\endgroup$
    – user106506
    Oct 3 '17 at 20:31
  • $\begingroup$ @user106506 read and familiarize yourself with Sobel's test. You cannot simply constrain the effect to be 0 with 0 standard error, or else everything is a mediator. Baron/Kenny is different. $\endgroup$
    – AdamO
    Oct 3 '17 at 21:06

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.