Hi Hui,

The PLS (variance-based) is very robust like covariance-based (MLE, GLS, ULS etc.).

There is a crescent tendency to use PLS because our data, in such cases, is free distribution, i. e., do not have multivariate normal distribuition.

See:

John J. Sosik, Surinder S. Kahai, Michael J. Piovoso (2009)

Silver Bullet or Voodoo Statistics? A Primer for Using the Partial Least Squares Data Analytic Technique in Group and Organization Research. Sage....

Get it at: http://journals.sagepub.com/doi/abs/10.1177/1059601108329198?journalCode=gomb

best regards

Dirceu

PLS-SEM is for usually for exploratory research designs. If you are doing conclusive research designs, first make sure that your theory is in place. It's better to go with SEM-AMOS if your research design is strongly supported by existing theories.