Latent variable without indicator

[Hansson]

Hi there,
I have little trouble with designing the path model within smartPLS 3.

My model contains, among other variables, three latent variables ("environmental attitude", "attitude towards wind energy", "attitude towards green investments") whitch are meassured by some indicators.
Now I want to reunite these three latent variables to just one latent variable called "attitudes".
When creating a this new latent variable "attitudes" smartPLS tells me to add an indicator to the variable but unfortunately there doesn´t exist one in my model because it was not possible to meassure such a complex attitude.

So my question is: Is it possible to reunite some latent variables into just one latent variable without using an indicator for the "new" latent variable?
Thank you very much for your support.

Best,
Hansson

[jmbecker]

All constructs or latent variables need indicators in PLS.
What you have in mind could be modeled by a higher-order construct. For this approach, you can either repeat the indicators of the lower-order constructs on the higher-order construct (e.g., attitude) or use a two-step approach with the latent variables scores from the lower-order constructs as indicators for the higher-order constructs.

Some literature on HCM:

Ringle, C. M., Sarstedt, M., and Straub, D. W. 2012. "A Critical Look at the Use of PLS-SEM in MIS Quarterly." MIS Quarterly 36 (1): iii-xiv.

Becker, J.-M., Klein, K., and 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.

Hair, J. F., Hult, G. T. M., Ringle, C. M., and Sarstedt, M. 2014. A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM). Thousand Oaks, CA: Sage.

However, with these constructs it is always the question, why you don't analyze the effect of each single constructs on your dvs and why it is advantageous to use a HOC. Some reviewers view second-order type model quite critically and advise to use to more complex model instead.

[RaffaeleP]

If you decide to operationalize your SEM with higher-order constructs than I would suggest this video

https://www.youtube.com/watch?v=kPeUTKjMF7o

[Hansson]

Thank you very much jmbecker and RaffaeleP for your support.
Indeed, higher-order constructs are what I am looking for.

@jmbecker: I want to use HOC to test Theory of Planned Bahavior (TpB) for investment decisions. Therefore, some latent variables have to reunite to the three constructs of TpB called "attitude", "subjective norm" and "perceived behavioral control". Hair, J. F., Hult, G. T. M., Ringle, C. M., and Sarstedt, M. 2014 recommend to use CB-SEM istead of PLS-SEM to test theories. However, I want to give smartPls a try on this one...

Best,
Johannes

[yaya]

Hello
Is this software, you can use the software Smart PLS
Factor
Instead marker or indicator
Use
please guide me

[JH Tan]

just wondering besides make this model become high order model, is there any other way?

such as get the mean of latent variable (environmental attitude", "attitude towards wind energy", "attitude towards green investments" in this case) and make it become indicator of "attitudes"(latent variable) to form a simple PLS-SEM model?