Multiple competing mediators with multiple IVs

[to67]

Hello,

I am assessing a multiple mediation model with 2 mediators (theoretically in competing nature) and multiple IVs. Following many key publications, such as "A Primer" by Hair et al. (2016), I assessed the model through bootstrapping and got total effects including direct effects, total indirect effects, & specific indirect effects (I am using SmartPLS 3.2.7). Total indirect effects and almost all specific indirect effects are significant, suggesting that mediation exists.

I then was comparing direct effects drawn from the total effect table against each of specific indirect effects to see types of mediations but struggling to figure out the process and interpretation. I've looked into many publications. Hair et al. (2016) has sections about multiple mediation models, but I couldn't find answers (Exhibit 7.8 presents a multiple mediation model, but I think the case study for mediation is about 1 mediator.). So appreciate your help.

A specific example is --
I included all mediators & IVs in one mediation model. Then, I got 1 direct effect (from a total effect table) whereas I have 2 specific indirect effects (from a specific indirect effect table).

Specific Indrect Effects
"Negative" Specific Indirect Effect from IV 1 --> Mediator 1 --> DV
"Positive" Specific indirect Effect from IV 1 --> Mediator 2 --> DV

Direct Effect
"Positive" Direct Effect from IV 1 --> DV

Question 1)
In this case, should I conclude IV1 -- Mediator 1 is competitive, whereas IV 1 -- Mediator 2 is complementary? So, could I say something like, Mediator 1 is a suppressor? Or, did I do wrong and should I get direct effects for each mediator (But, I think I should not run a different model with only one of these mediators, following the literature I've read.).

Question 2)
A specific indirect effect table shows that specific indirect effects from most IV 1 via Mediator 1 are significant and in the hypothesized direction (negative), whereas as above shown, a direct effect from IV 1 --> DV is positive (competitive mediation). In this case, could I say the hypotheses were supported while we should investigate more of the mechanism (such as an additional or a different mediator, or moderator, etc.) in future research?

Thank you so much!

[jmbecker]

Inclunding all mediators in one model is correct.

Q1
I would then assess each mediator separately.
First: Is the sign of the specific indirect effect plausible?
Second: Compared to the direct effect is it complementary or competetive?

BTW: You did not mention, whether your direct effect is (IV 1 --> DV) is significant. If not, you have full mediation and the sign of the effect is irrelevant.

Q2
Yes, that sounds correct.