Hello.
I have a big problem in understanding a reviewer's comment. The reviewer says:
"One particular concern, often overlooked in PLS papers, is the lack of explicit convergence between the outcomes of the factor-analysis and the final model. Clearly, and by definition, the cross-loadings in the measurement model show up as correlations in the measurement model. It is almost impossible to disentangle the two, but their interpretations are quite different. Please give this some attention"
I do not understand this comment at all and I have no clues on how to handle it. I also read the new book by Hair et al. but I found no reference to such issue. Do you have any clues? What does the reviewer mean? Is there any published paper/book addressing (and hopefully explaining better) such issue?
Thank you very very much for your help ...
Irene
unclear reviewer's comment on cross-loadings vs correlations
-
- PLS Junior User
- Posts: 10
- Joined: Mon Mar 26, 2012 3:06 pm
- Real name and title:
- Location: Università Bocconi
- Diogenes
- PLS Super-Expert
- Posts: 899
- Joined: Sat Oct 15, 2005 5:13 pm
- Real name and title:
- Location: São Paulo - BRAZIL
- Contact:
Some comments
"One particular concern, often overlooked in PLS papers, is the lack of explicit convergence between the outcomes of the factor-analysis and the final model.
1) Bido: He is not clear, maybe he is talking about run a confirmatory factor analysis before the structural model (two step approach), or have you run a exploratory factor analysis before PLS-PM?
If you want to run a CFA using SmartPLS:
Connect all LV one with other, take care to avoid feedbacks.
Run the PLS algorithm with “factor” weighting scheme
Assess just the measurement model:
Convergent validity: outer loadings, AVE
Reliability: composite reliability, Cronbach’s alpha
Discriminant validity: indicator level = crossloadings and LV level = √AVE>r_VL
Clearly, and by definition, the cross-loadings in the measurement model show up as correlations in the measurement model.
2) Bido: Ok.
It is almost impossible to disentangle the two, but their interpretations are quite different. Please give this some attention"
3) Bido: What “two” things is he talking about?
I suggest you, before invest your time trying to correct something, it is better to check with editor, if it is possible to improve the explanation of these issues.
One thing that I could imagine, but it is difficult to figure out what he was thinking:
- Outer loadings, in fact, are “standardized regression coefficients”, but when the factors in an EFA are orthogonal (independent), the standardized regression coefficients are equal correlations.
- When the factors in EFA are correlated (oblique rotation) we should not interpret outer loadings as correlations, even they are being pretty similar to standardized regression coefficients.
- In PLS-PM the LV are correlated, for this reason, we should interpret outer loadings as standardized regression coefficients, but I am not sure if this is the issue that the reviewer was talking about.
Best regards,
Bido
"One particular concern, often overlooked in PLS papers, is the lack of explicit convergence between the outcomes of the factor-analysis and the final model.
1) Bido: He is not clear, maybe he is talking about run a confirmatory factor analysis before the structural model (two step approach), or have you run a exploratory factor analysis before PLS-PM?
If you want to run a CFA using SmartPLS:
Connect all LV one with other, take care to avoid feedbacks.
Run the PLS algorithm with “factor” weighting scheme
Assess just the measurement model:
Convergent validity: outer loadings, AVE
Reliability: composite reliability, Cronbach’s alpha
Discriminant validity: indicator level = crossloadings and LV level = √AVE>r_VL
Clearly, and by definition, the cross-loadings in the measurement model show up as correlations in the measurement model.
2) Bido: Ok.
It is almost impossible to disentangle the two, but their interpretations are quite different. Please give this some attention"
3) Bido: What “two” things is he talking about?
I suggest you, before invest your time trying to correct something, it is better to check with editor, if it is possible to improve the explanation of these issues.
One thing that I could imagine, but it is difficult to figure out what he was thinking:
- Outer loadings, in fact, are “standardized regression coefficients”, but when the factors in an EFA are orthogonal (independent), the standardized regression coefficients are equal correlations.
- When the factors in EFA are correlated (oblique rotation) we should not interpret outer loadings as correlations, even they are being pretty similar to standardized regression coefficients.
- In PLS-PM the LV are correlated, for this reason, we should interpret outer loadings as standardized regression coefficients, but I am not sure if this is the issue that the reviewer was talking about.
Best regards,
Bido