Ha Hoang, CB-SEM and PLS-SEM serve different purposes. Briefly put, you would use the former for testing theory (confirmatory) and the latter for developing theory (exploratory). However, if you use CB-SEM and encounter convergence issues, you can consider using consistent PLS-SEM as a valid alternative.

Dear Ha Hoang

The covariance-bases structural equation modelling (CB-SEM) requires normal-distributed input data which implies that it needs larger data samples to run appropriately, however, the partial least square structural equation modelling (PLS-SEM) does not imply restriction for the data distribution thus allowing it to run with lower sample sizes compared to (CB-SEM).

Another difference is that if the study includes a formative construct then the (CB-SEM) is not the tool to use to develop this model. Although that (PLS-SEM) is often viewed as a useful approach for prediction purposes, however, it is still not that developed yet to be predictive as other prediction well known statistical techniques.

With my experience, I could tell that using (PLS-SEM) tends to show better results since it more challenging to collect more data to follow the data distribution restriction of the (CB-SEM).

If you are interested to use it can be applied within the R environment via the following packages:

* (CB-SEM) is also available withing R using Laavan Package.

1. semPLS

2. plspm

Here is a book that was written by the designer of the (plspm) package to illustrate how to use it.

https://www.gastonsanchez.com/PLS_Path_Modeling_with_R.pdf

Also, you can use the SmartPLS software, which is a very powerful software that is mainly dedicated to run the partial least square analysis.

Finally, I would like to highlight the researchers’ arguments for choosing PLS as the statistical means for testing structural equation models (Urbach & Ahleman, 2010) to give you more insight into it:

• PLS makes fewer demands regarding sample size than other methods.
• PLS can be applied to complex structural equation models with a large number of constructs.
• PLS does not require normal-distributed input data.
• PLS is able to handle both reflective and formative constructs.
• PLS is better suited for theory development than for theory testing.
• PLS is especially useful for prediction.
• PLS can handle first and second-order together.
• Based on Chin, (1998a): In the latter case, PLS is used to develop propositions by exploring the relationships between variables.

Article Estimation issues with PLS and CBSEM: Where the bias lies!

Article PLS-SEM or CB-SEM: updated guidelines on which method to use

@ Mehmet Mehmetoglu thank you for your answer, it really help me to make clear between these two methods!

@Belal Edries, thank you so much for your valuable answer!