Mediation effect, rejection?

[Jako]

When do I have to reject the hypothesis of a Mediation effect?



As you can see, the path between DI and OP is relatively low and moderately significant.

The rest Shows a complementary partial Mediation effect. But is the path between DI and OP sufficient?

[jmbecker]

Important is the indirect effect from OA on OP (the product of the effect of OA on DI * DI on OP). If that indirect effect is significant you have a mediation. If it is not significant you don't have mediation.
Please also search the forum. There are a lot of discussions about medition. Particularly, I would recommend to read the article by Zhao et al.

[Jako]

Dear Dr. Becker,

thanks for the reply. I have read all the Forum Posts and it didn't really help.

The first thing I always need to prove is the significance of the indirect effect. The rest is safe. Exactly this first step is what I'm unsure of.

I have one very significant path (OA->DI) and one moderately (t>=1,645; t<=1,96) significant. Does this mean, that my indirect effect is significant? a*b is 0.137; (->VAF 0,23, knowing that ist not important any longer)

Moreover, it is recommended in Zhao et al. to examine the lower and upper bounds of the indirect effect. If I check this in SmartPLS 3 after Bootstrapping > Indirect Effect > Bias Corrected
I get the following Picture:



OA -> OP indirect effect is below 0 now. Does that force me to reject my Mediation?



Here are the total effects.

Sorry, if I misunderstood some explanations. Even through carefully reading the book Primer on PLS-SEM I cant find an answer.

[Jako]

Addition: I am using a hierarchical model so I implemented a two-stage Approach.

[jmbecker]

Your indirect effect seems to be not significant (at a 5%) level, because the confidence interval includes zero (even though the lower bound is only slighly below zero). It would probably be significant at a 10% level though. Hence, you cannot conclude a mediation at a 5% error probability level.