I have a research design with 3 predictors and 2 mediators, tested on 2 different samples with multiple regression analysis. For my first sample (N=70) the results were:variables(mediators) entered at the second step predicted significantly the outcome, but only one of them had a significant coefficient. The IV's had non significant coefficients at step 2, and the beta's decreased in size. In my second sample (N=100), same set of variables, at step 2 the proposed mediators didn't predict significantly the outcome and none had significant coefficients. The IV's also had non significant coefficients at step 2 and they decreased in size. My question - Is mediation supported for the first sample? What about my second sample, what can I say about the results if none of the mediators was significant, can I draw a valid conclusion and still use a mediational design?

All the variables are positivelly correlated and there may be some multicollinearity problems for one of the mediators in each study. Some of the predictors have high correlations with the mediators, above .6.

  • $\begingroup$ It's hard to say based on what is listed here, but one possibility is that you have insufficient statistical power. $\endgroup$ – gung - Reinstate Monica Jul 22 '12 at 19:14

I have a comment and hopefully an answer.

You use the term "second step", which is a term typically reserved for hierarchical models (e.g., hierarchical regression). Are you certain you are doing mediation analysis? It may help if you describe your question and perhaps how you performed your analyses (e.g., using path analysis, a macro in SPSS). Note that since you have two mediations in your proposed model, it sounds like you would be doing a multi-mediational model, which is something relatively advanced which would likely require path analysis, a programming language, or a macro.

I think the answer to your question is "no", assuming you are doing meditational analysis. For a mediation to be significant, you would need to have a significant direct effect between your mediator and DV, and a significant effect between your IV and mediator, at the very least and in most cases. It doesn't sound like you have this. For a mediation to be significant, your IV needs to "cause" a mediator to such an extent that the mediator's effect on the DV can be at least partially attributed to your IV.

Here is some good basic information on mediation:


  • $\begingroup$ Thank you for the reply. I am doing the analysis in SPSS. Yes, it is a hierarchical model, that is what I understood from some sources, ex. www2.psy.uq.edu.au/~wlouis/stats/mediationv4.doc Tested separatelly, the significant effects between the terms exist. Would it be better to test the effect of each mediator so I wouldn't have multiple mediation? it gives me significant results.. $\endgroup$ – andree Jul 22 '12 at 20:48
  • $\begingroup$ Ah, I see. Well I don't think thats a very efficient or proper way to do a mediation analysis, but it might get the job done. You will not be able to interpret specific paths in your model (e.g., through which mediator the effect is occurring), but it will tell you if variance between your IV and DV is shared by other variables. If this is for class, then try your best to do what the instructor wants. If it is not for class, then take a look at this wonderful macro which works in SPSS: afhayes.com/spss-sas-and-mplus-macros-and-code.html $\endgroup$ – Behacad Jul 22 '12 at 20:51
  • $\begingroup$ I foundd the macro but it doesn't seem to work with my older version of spss 16, keeps giving me errors. I could try to follow the steps described in some of these papers and use the sobel test to calculate the effect of each mediator holding the other m constant a.s.f. Still, the question was referring to the insignificant coefficient for my mediators. Can I test the for mediation in this case? $\endgroup$ – andree Jul 23 '12 at 7:22
  • $\begingroup$ If your mediators are not significantly correlated with your dependent variable, then no, there is no mediation. That being said, it might be because you are testing them in conjunction. Try doing Sobel's test and the steps described by Kenny using only one mediator at a time. $\endgroup$ – Behacad Jul 23 '12 at 13:01
  • $\begingroup$ They are significantly correlated. I tried calculated Sobel for a mediator at a time and there is mediation. The problem was that I read a similar article where both mediators were tested in the same time and I tought the question would come why I didn't test for both mediators in the same time and did separate analysis... So they must be significantly correlated with the criterion, that all, it doesn't matter if they are significant in regression? maybe the macro will work.. Thanks much for answering. $\endgroup$ – andree Jul 23 '12 at 15:29


For a mediation to be significant, you would need to have a significant direct effect between your mediator and DV, and a significant effect between your IV and mediator


This is not necessary for contemporary mediation analysis


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.