Mediation testing

This forum is closed, and read-only.
Locked
Hedwigga
PLS Junior User
Posts: 2
Joined: Fri Dec 13, 2019 9:58 pm
Real name and title: N. Hamdi

Mediation testing

Post by Hedwigga »

Hello. I am a beginner in SmartPLS and now using smartpls 3.2.9. I have been reading Zhao 2010, Hair 2017, Memon 2018, and several more papers. But since I'm new to both statistics and pls sem, I get easily confused. In my model, I have 3 IVs going through each of 4 MVs into 1 DV as below:

IV1 --> MV1 --> DV
IV1 --> MV2 --> DV
IV1 --> MV3 --> DV
IV1 --> MV4 --> DV

IV2 --> MV1 --> DV
IV2 --> MV2 --> DV
IV2 --> MV3 --> DV
IV2 --> MV4 --> DV

IV3 --> MV1 --> DV
IV3 --> MV2 --> DV
IV3 --> MV3 --> DV
IV3 --> MV4 --> DV

My questions regarding the test are:

1) When i draw the path model in smartpls3, should i connect
IV1 --> DV,
IV2 -- > DV,
IV3 --> DV too?

2) What is the difference between total indirect effect and specific indirect effect? If specific indirect effect are mostly not significant (pvalue), but total indirect effect is significant, does it means there is mediation or not?

3) I read about decision tree in Zhao 2010. I understand that a&b are indirect effects while c is direct effect. ( a = IV -> MV, b= MV -> DV, and c = IV -> DV). However, which value is actually a, b and c? Are the values are the original sample?

4) How to get direct effect? I understand total effect = total indirect + direct effect. Which value should i minus to get direct effect?

I hope i can get answers. I've been watching youtube videos too but I don't quite understand. Thank you..
jmbecker
SmartPLS Developer
Posts: 1281
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: Mediation testing

Post by jmbecker »

1) If you want to test a proper mediation, you would need to connect the IV and DV too.

2) Yes and no. If the specific indirect effect is not significant, it means that the mediation through a specific mediator cannot be concluded.
The total indirect effect is the combined indirect effect through all mediators at the same time. It makes only sense to analyse this if you also have the direct effect (IV --> DV) in place. It might indicate that there is some form of mediation, but each separate mediation might be too weak.

3) In the bootstrapping these are the original sample coefficients, yes.

4) You get the direct effect from IV --> DV in a model where the mediator MV is also included.
Dr. Jan-Michael Becker, BI Norwegian Business School, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
Locked