If you use composite models, you may just want to use PLS; also PLSc can be used to estimate a model that uses both common factor models and composite models. For the decision, which proxy you would use for a latent variable (composite or common factor), you may find this article useful:
http://www.sciencedirect.com/science/ar ... 6316304404
If you like to estimate factor models, use CB-SEM. Here you need a good fit of the model (
https://www.smartpls.com/documentation/ ... /model-fit). However, you can mimic common factor models (and CB-SEM) by using PLSc:
https://www.smartpls.com/documentation/ ... ith-cb-sem
However, the usual question that arises: Why would you like to mimic factor models when you can use CB-SEM. Also, when using PLSc to mimick factor models, you need to ensure a good fit for your model. Is the SRMR <0.08 after running PLSc? If the fit is not good, PLSc provides the strange results that you obtained (PLSc and its bootstrapping is very vulnerable to constructs with low reliabilities, e.g., Cronbach's alpha below 0.8).
When you consider composites, you can use the usual PLS-SEM algorithm. A good start to addressing this may be to look into composites as proxies of the conceptual variables of interest. Here are some article that you may find useful:
- Sarstedt, M., Hair, J. F., Ringle, C. M., Thiele, K. O., & Gudergan, S. P. (2016). Estimation Issues with PLS and CBSEM: Where the Bias Lies! Journal of Business Research, 69(10), 3998-4010. http://www.sciencedirect.com/science/ar ... 6316304404
Rigdon, E. E. (2012). Rethinking Partial Least Squares Path Modeling: In Praise of Simple Methods. Long Range Planning, 45(5-6), 341-358. http://www.sciencedirect.com/science/ar ... 0112000581
Rigdon, E. E. (2014). Rethinking Partial Least Squares Path Modeling: Breaking Chains and Forging Ahead. Long Range Planning, 47(3), 161-167. http://www.sciencedirect.com/science/ar ... 0114000144
In addition, here you find some argument when to use PLS-SEM:
Richter, N. F., Cepeda Carrión, G., Roldán, J. L., & Ringle, C. M. (2016). European Management Research Using Partial Least Squares Structural Equation Modeling (PLS-SEM): Editorial. European Management Journal, 34(6), 589-597.
http://www.sciencedirect.com/science/ar ... 7316300925
And some argument that you don't want to use:
Rigdon, E. E. (2016). Choosing PLS Path Modeling as Analytical Method in European Management Research: A Realist Perspective. European Management Journal, 34(6), 598-605.
http://www.sciencedirect.com/science/ar ... 7316300585
Also, when using composites and PLS-SEM, you may want to assess the predictive quality of the model by using the Blindfolding procedure and reporting the resulting Q² values as well as applying the PLSpredict procedure (which is implemented in SmartPLS). Here is some more info about PLSpredict:
https://www.smartpls.com/documentation/ ... es/predict
We hope that you find this ideas helpful!