No convergent validity at all - help! i need somebody...

[HansDampf3000]

Hello,

i conducted a survey a few month ago. My boss told me to do an sem analysis according to the opportunity ability and motivation theory.
So i aligned the matching itmes to these predictors in order to check what forms the intention of my survey participants (yes i am going to / no i am not going to).
Because of the fact that i aligned the items afterwards because of what literature says what ability or opportunity is, these predictors are very inconsistent i think.
e.g.

Opportunity are the items: structural complexity, hart to find craftsmen, legal requirements are complex, no time planning, no time conducting, no objective information (i totally agree /...../i dont agree at all)
Ability: I would have problems financing, house is demanding, friends and family find ... good, neighbourhood tinks... is good, my age is a barriere, i am emotionally attached to my house...
Furthermore my Motivation-predictor is a calculated value and ranges from -1 to + 3
(all my constructs seem to be formative to me)

I also included some environmental statements after Bearden et al 2011 (reflective).
When i tried to checl the convergent validity all values were really bad, eg. 0,15 or 0.2

What can i do in order to improve my SEM?
I would really appreciate your help! thank you!

Sincerely,
HansDampf3000

[jmbecker]

How are you checking for convergent validity of your (obviously) formative constructs?

[HansDampf3000]

How are you checking for convergent validity of your (obviously) formative constructs?

Dear Dr. Becker,
i used the following statements (5-point likert scale) and directed the affiliated reflective construct to the formative constructs (one after another).
Do i have to sort out bad items (low loadings/weights/non significant) before or after this test?

1. It is important to me that the products I use do not harm the environment.
2. I consider the potential environmental impact where the original scaleof my actions when making many of my decisions.
3. My purchase habits are affected by my concern for our environment.
4. I am concerned about wasting the resources of our planet
5. I would describe myself ashas a high level of reported
6. I am willing to be inconvenienced in ordervalue to take actions that are more environmentally friendly.

Do you think it would be better to use only some of these questions for the convergent validity test, e.g. 6,3 and 1, which fit - in my humble opinion - good to each other?

Thank you very much!

kind regards,

Hans Dampf

[HansDampf3000]

Ok, i think i just found the reason for the bad results.
My model consists of 3 predictors (Motivation, Opportunity, Ability; all formative) and one dependent variable (Investment sum).
To use an additional reflective construct to check the convergent validity and remove it afterwars was the mistake. I just misunderstood this part in Hair et al 2017...
However, is there something else I can do to satisfy reviewers? My colleagues, who did the survey/data i use right now, told me they used another book for preparation (Huber, ..., 2007) of the SEM. Within this book it is stated that convergent-validity cannot be checked for formative constructs.

After reducing the items with low quality (following to Hair et al 2017) from initial 29 only 13 items were left. 3 belong to Opportunity (2 significant), 3 to Ability (2 significant) and 7 to Motivation (2 significant; 1 negative weight, 1 positive weight). All three path coefficients are significant.
R^2 = 0,24 (according to Hair this is good for consumer behavior into which i am investigating), f^2s are 0,1; 0,12; 0,03

Are these results worthy for publishing?

Thank your for your help!

Kind regards,

HansDampf3000

[jmbecker]

If you don't have a corresponding reflective construct for your formative constructs you cannot check convergent validity as described in Hair et al by means of the redundancy analysis.
Another option would be the inclusions of an established reflective construct where literature has shown very high correlation with (e.g., if you are evaluating satisfaction, the satisfaction-loyalty link is very established). Such a construct is however often hard to find and/or to justify.

If you don't have any such a construct in your data there is no way to check convergent validity.

You are left with face validity: do your predictors relate to your dependent variable in an expectable way? Are those formative items significant that one would expect to have the strongest influence given your research model?
Of course, all of these questions are very vague and hard to defend.