PLS versus CFA

[svenaik]

Hi

I recently read that PCA (principal components analysis) is appropriate methodology for formative models, wheres CFA (confirmatory factor analysis) is the appropriate method to use for reflective models.

Since PLS is based on PCA and not CFA, is it correct to use PLS to estimate reflective models. If yes, how does one reconcile with the view above.

Thanks in advance for your opinions/suggestions/comments.

Best

Sunil

[viswadatta]

Yes PLS can be used with both reflective and formative models. Covariance methods use only reflective indicators. PLS is also confirmatory in nature and can be uced to test larger, more complex models with smaller sample sizes.

[Diogenes]

Hi,

a great discussion in:
WESTLAND, J. C. Confirmatory Analysis with PLS. april 2007.
available in: http://www.rhsmith.umd.edu/dit/events/p ... lpaper.pdf

Best regards.

Bido

[bwilson]

Dear all,

I think the PLS community need to not use blanket statements that CBSEM cannot be used for formative modeling.

It has numerous difficultlies:
identification
programming
convergence, etc

but it can be done.

PLS offers more flexibility in the general sense of the word when modeling with formative measures.

Brad