SmartPLS EFA

[krismurphy01]

Hello,

Let me begin by stating I am committed to using SmartPLS and the product is superior in many ways to SPSS/AMOS.

Now that I have been setup on this board I would like to share some information on EFA in SmartPLS versus SPSS. The challenge that I have is running an EFA in SPSS yields substantially lower cross-loadings than SmartPLS. Technically, I am running the EFA process using SmartPLS by way of a SEM model - per the Hair book. I have 24 indicators spread across each of my 8 constructs with a sample size of 158 - an elite group.

Given the nomological net comment below maybe a CFA is superior to EFA. On the other hand, I was able to identify some sub-dimensions through the EFA process so it does add some value.

Are there any EFA tricks in SmartPLS that others would like to share? My construct CFA and SEM validity and goodness of fit metrics are good.

Response from SmartPLS support team - thank you.

It is important to understand that PLS and EFA produce different scores. Both methods different! PLS produces latent variable scores considering the nomological net. Whenever you change your path model, your scores and, thus, the outer loadings change. But this make total sense since estimating the scores considering the context is the purpose of PLS.

You only get the same score compares with SPSS in SmartPLS if you to the following.
•Use two constructs A and B
•Create a relationship between A and B
•Assign exactly the same indicators to A and to B
•Run the PLS algorithm
The results give you a structural relation ship of 1 and an R² of 1 for the dependent constructs. The results are irrelevant. The outer loadings are identical in both constructs and represent factor loadings.
These factor loadings are identical to PCA (certainly no rotation) using SPSS.

All other results in SPSS ale always different.