forum.smartpls.com

Hi PLS-Experts and Users! I have been using Smart-PLS since one year. The purpose is to set-up and empirically test a model about Transaction Success in cross-border acquisitions of German Companies.
After setting up the hypothesis, collecting manifest variables, I set up the formative model as follows:

[i

The dependent variable is logically a reflexive model...

So, my question is, how to interpret the manifest variables and the constructs in that case? After that step I eliminated few variables and achieved slightly better t-values.
The problem is also a small sample size of 45 cases. Are there to many variables for it?
The VIF values, before and after variable selection are all below 2.

I would be thankful for any help...
My best...
Al

Hi,

First, I think the sample 45 is not enough to execute this model. Please check the adequacy of your sample with G*Power.
Second, you've got a problem on the sign of formative indicators, which partially positive and the other negative. It shows that there is a problem on your questionnaire. And there is also an indicator that you are not significant to change the meaning of the construct.

Greetings,

Hi,
thanks for G* ;-) I will check it.

Could you pls. specify the other points: what do you mean by "not significant to change the meaning of the construct"?
What would you suggest in that situation - dropping the variables? There are manifest variables in the theoretical literature, which are obviously not that important for the practictioners (in this case: risk factors).
Regards

no further comments, opinions or suggestions?

Hi,

You can drop a formative indicator only with the support of the theory, or you can keep it only if the outer loading is pretty good (> 0.4). Formative indicators are expected to only slightly used, because it will cause a lot of problems if > 3 items.

Greetings,

Hi Hengkov,
could you pls. look the changed model - theoretically the decrease to 3 constructs is fine.
Yet, I still have variables with negative signs - what do you think about this - does each manifest variable needs positive weight?
BTW: What are reliable path values? Over 0.4/0.5?