Add relations to lower-order constructs

[mnabi]

Hi everyone,

Please I need your help.

I don't know how to test my model. The problem is about testing a construct with two dimensions. I have a formative second-order construct (success) and I want to add relationships among independent variables with these two dimensions (lower-order constructs). You can see my model in picture below. Is it possible in Smart PLS and how?

Success Model.JPG (39.11 KiB) Viewed 7969 times

Best
Morteza

[agalvez]

Hi mnabi

I'll try to help you but I'm a newbie in PLS.

If I'm not mistaken, you have to create two construct for both variables: presence and participation. Then you have to "path" these two to the second order construct, named Success. Let's suppose that Presence has three items (PR1, PR2 and PR3) and Participation has also three items (PA1, PA2, PA3). Just add this items to each construct and all of them (from PR1 to PA3) to the second order construct, Success.

I've read some papers about path modeling and I think this is the correct procedure to build a 2nd order construct.

[jmbecker]

I would not model this as a second-order constructs, but simply use both constructs (Presence & Participation) as separate concepts in your model. You can then draw paths to both variables and assess the impact of your IVs on the two success measures.

[agalvez]

I would not model this as a second-order constructs, but simply use both constructs (Presence & Participation) as separate concepts in your model. You can then draw paths to both variables and assess the impact of your IVs on the two success measures.

Yes. I think this allows you to have more information about the impact of IV on DP.

As far as I know, 2nd order constructs are used, i.e, in complex models with a lot of structural paths. In this case, maybe is better to use only 1st.

Regards.

[mnabi]

I would not model this as a second-order constructs, but simply use both constructs (Presence & Participation) as separate concepts in your model. You can then draw paths to both variables and assess the impact of your IVs on the two success measures.

Thank you very much.
1. This is a good solution. However, in this solution, I can assess the two constructs (Presence and Participation) that form the Success construct. Now, how can I measure the Success construct itself?
In other words, in the higher component model of Success, I can use the repeated indicators approach to measure Success, but In your solution, I should separate the Success construct from its formative constructs (Presence and Participative). Is there anything that I don't understand?

2. Hair et al. (2017) mentioned that the number of indicators should be similars across the lower order constructs in the repeated indicators approach. But, What if they won't be similar? Should I eliminate some indicators or is there another approach?

[agalvez]

I would not model this as a second-order constructs, but simply use both constructs (Presence & Participation) as separate concepts in your model. You can then draw paths to both variables and assess the impact of your IVs on the two success measures.

Thank you very much.
1. This is a good solution. However, in this solution, I can assess the two constructs (Presence and Participation) that form the Success construct. Now, how can I measure the Success construct itself?
In other words, in the higher component model of Success, I can use the repeated indicators approach to measure Success, but In your solution, I should separate the Success construct from its formative constructs (Presence and Participative). Is there anything that I don't understand?

2. Hair et al. (2017) mentioned that the number of indicators should be similars across the lower order constructs in the repeated indicators approach. But, What if they won't be similar? Should I eliminate some indicators or is there another approach?

1. I think that if you separate the construct, you refer to Success when talking about this two factors: Presence and Participation. Maybe I'm mistaken, but the only way to measure ""Success"" itself is with 2nd order (CR, AVE, etc.). With 1st order you refer "Success" indirecly trought its two constructs, that allows you to evaluate the impact of each of then on IVs.

[mnabi]

I would not model this as a second-order constructs, but simply use both constructs (Presence & Participation) as separate concepts in your model. You can then draw paths to both variables and assess the impact of your IVs on the two success measures.

Thank you very much.
1. This is a good solution. However, in this solution, I can assess the two constructs (Presence and Participation) that form the Success construct. Now, how can I measure the Success construct itself?
In other words, in the higher component model of Success, I can use the repeated indicators approach to measure Success, but In your solution, I should separate the Success construct from its formative constructs (Presence and Participative). Is there anything that I don't understand?

2. Hair et al. (2017) mentioned that the number of indicators should be similars across the lower order constructs in the repeated indicators approach. But, What if they won't be similar? Should I eliminate some indicators or is there another approach?

1. I think that if you separate the construct, you refer to Success when talking about this two factors: Presence and Participation. Maybe I'm mistaken, but the only way to measure ""Success"" itself is with 2nd order (CR, AVE, etc.). With 1st order you refer "Success" indirecly trought its two constructs, that allows you to evaluate the impact of each of then on IVs.

With many thanks

For measuring "Success", I found two-stage approach PLS-SEM Models with unequal numbers of indicators in one of Dr. Becker papers. However, I don't know whether I can use it here or not. According to two-stage approach, for first stage, I measure my model without the "success" construct and report that:

First stage.JPG (39.8 KiB) Viewed 7824 times

And for measuring "Success" (especially R2), I use second stage like this:

Second stage.JPG (45.19 KiB) Viewed 7824 times

Can I use this approach like above? Any idea?

Best,

Becker, J. M., Klein, K., & Wetzels, M., (2012), "Hierarchical Latent Variable Models in PLS-SEM: Guidelines for Using Reflective-Formative Type Models", Long Range Planning, 45(5–6), 359–394. http://doi.org/10.1016/j.lrp.2012.10.001