I´m performing a mediation analysis. I have one predictor variable (dychotomous), three mediator variables and one outcome variable. I´m not sure wich code should I use to this analysys, and also how to perform the sobel test (is there a need for a sobel test when performing mediation analysis using SEM?).
Until now, i´ve been using the following code:
MODEL: cAE ON contint; (mediator1 ON predictor) cAS ON contint; (mediator2 ON predictor) cER ON contint; (mediator3 ON predictor) cTL ON cAE cAS cER;(outcome ON mediators) cTL ON contint; (outcome ON predictor).
Is there anything missing in this code? What else should I take in account?
All my p-values are significant, but I´m not sure if I can affirm that I prooved the mediation model just like that.
A good source for mediation is the 2013 Hayes book. Mediation usually involves looking at indirect effects. See Example 3.16 in the user's guide. I would recommend CINTERVAL (BCBOOTSTRAP) instead of CINTERVAL which is shown in this example. For further information, you can look up these options in the user's guide.