Hello,
Suppose there are five independent variables and one dependent variable in a model. All the five independent variables are hypothesized to have direct influence on the dependent variable. However, one of the independent variables has also been hypothesized to be a mediator between another independent variable and the dependent one. Now, we want to investigate the relative influence of the independent variables on the dependent variable.
Q) Should the structural model be tested without mediation? or with mediation? (objective is to check relative influence of the independent variables only); If with mediation is required, then should the the relative influence be checked on the basis of path coefficients (direct effect)? or on the basis of total effects?
Thanks
Relative significance of exogenous variables

 SmartPLS Developer
 Posts: 1094
 Joined: Tue Mar 28, 2006 11:09 am
 Real name and title: Dr. JanMichael Becker
Re: Relative significance of exogenous variables
It depends how you define the relative impact in a mediator model. It is not as straightforward and clear (especially to disentangle the impact of the mediator). You would certainly want to look at total effects there. If you look only at direct effect you would get the same results as when not doing the mediation (if your mediator would be a normal predictor).
Dr. JanMichael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
Re: Relative significance of exogenous variables
Suppose, 5 independent variables are a, b, c, d, and e and the dependent variable is y. I want to check relative influence of 5 IVs on the dependent one. However, I want to check whether 'a' (IV) also acts as a mediator between 'b' (IV) and y (DV).
Case 1: when I test structural model without mediation
path a>y exhibits highest path coefficient followed by b, c, d, e
Case 2: when I test structural model with mediation
model exhibits partial complementary mediation
path a>y still exhibits highest path coefficient followed by c, b, d, e (direct effect of b gets lower than c); now, total effect of b is higher than a.
I have two objectives:
i) To examine relative influences of IVs on the DV.
ii) To test the mediation effect.
Which of the following options is suitable? please suggest
1) I should go for only one structural model (with mediation) and calculate relative influence on the basis of total effect.
2) I should go for only one structural model (with mediation) and calculate relative influence on the basis of path coefficients?
3) I should go for two separate structural models? first, one without mediation and calculate relative influence on the basis of path coefficients; second, test the mediation.
Case 1: when I test structural model without mediation
path a>y exhibits highest path coefficient followed by b, c, d, e
Case 2: when I test structural model with mediation
model exhibits partial complementary mediation
path a>y still exhibits highest path coefficient followed by c, b, d, e (direct effect of b gets lower than c); now, total effect of b is higher than a.
I have two objectives:
i) To examine relative influences of IVs on the DV.
ii) To test the mediation effect.
Which of the following options is suitable? please suggest
1) I should go for only one structural model (with mediation) and calculate relative influence on the basis of total effect.
2) I should go for only one structural model (with mediation) and calculate relative influence on the basis of path coefficients?
3) I should go for two separate structural models? first, one without mediation and calculate relative influence on the basis of path coefficients; second, test the mediation.

 SmartPLS Developer
 Posts: 1094
 Joined: Tue Mar 28, 2006 11:09 am
 Real name and title: Dr. JanMichael Becker
Re: Relative significance of exogenous variables
I would probably do no. 3 (i.e., both models) and also discuss the difference.
However, it is strange the the direct path estimates on Y change in their order and magnitude when you include a path from b to a. All the rest of the model should be the same and therefore also not affected by this inclusion (unless PLS model estimates have not be robust before).
However, it is strange the the direct path estimates on Y change in their order and magnitude when you include a path from b to a. All the rest of the model should be the same and therefore also not affected by this inclusion (unless PLS model estimates have not be robust before).
Dr. JanMichael Becker, University of Cologne, SmartPLS Developer
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de
Researchgate: https://www.researchgate.net/profile/Jan_Michael_Becker
GoogleScholar: http://scholar.google.de/citations?user ... AAAJ&hl=de