Confirmatory Composite Analysis (CCA): Assessing quality of measurement model in SmartPLS (PLS-SEM)

This video illustrates a method "#ConfirmatoryCompositeAnalysis (CCA)" proposed by Hair, Howard, and Nitzl (2020) to assess quality of #MeasurementModel when using PLS-SEM for structural equation modeling in #SmartPLS. CCA proposes inclusion of #Nomological and #PredictiveValidity into the procedure of assessing quality of measurement model when using PLS-SEM. CCA is an alternative approach to #ConfirmatoryFactorAnalysis (CFA) we use frequently to assess quality of measurement model in CB-SEM using #AMOS software. Confirmatory Composite Analysis (CCA) can be conducted by following 7 steps approach to assess reflective model, and 5 steps approach to assess formative model.
Reference: Hair, Howard, and Nitzl (2020). Assessing measurement model quality in PLS-SEM using confirmatory composite analysis. Journal of Business Research Volume, 109, 101-110. Available on:
www.sciencedir...
You can join our scholarly social network Scholars Index by visiting our site:
www.scholarsind...

[@tati29991]

In most of the works that use PLS-SEM, they start from the fact that the instrument is already validated. However, I have had many doubts in the case where a new instrument is designed (based on theory and adjusting scales). The question is whether, prior to the evaluation of the measurement and structural models, the instrument must be validated. And if so, how do I do it when I have a small sample? Or if, on the contrary, the evaluation of the measurement model serves to justify the validity of the instrument. Thank you for your attention.

[@ChMahmoodAnwar]

Although SEM or PLS factor loadings are considered as equal to traditional Factor Analysis loadings by many scholars, I don't withstand with this notion. I recommend to input Pattern Matrix results into SEM or PLS model to get fully validated outcome.

[@angieginger1690]

Is it possible if you could do a practical tutorial video on how to use smartpls to conduct CCA? Thanks

[@TestTest2022TestTest]

Hello! Thanks for the great video! Is there a suggestion to use factor instead of path as the weighting scheme for CCA? I am just unsure cause Hair et al. (a primer on partial least squares structural equation modeling) use path to report statistics like factor loadings or AVE. Maybe you could give a quick statement if both is correct.

[@ChMahmoodAnwar]

If your objective is to test the quality of Measurement Model, you should select Factor. Otherwise, path.

[@amaliks13]

Hi, have you made the 7 step and 5 step pictures yourself or taken from the book? I want to give reference to these pictures in my research work. Can you please guide how to refer?

[@ChMahmoodAnwar]

The CCA steps were extracted from Hair et al. (2020). I depicted their flow. You can cite videos as per APA video citation style.

[@priyadarshinilaishram4206]

Sir , which software tool should be used for doing CCA ?

[@ChMahmoodAnwar]

CCA is a manual process you need to separately work on each step. Theoretically, these steps can't be incorporated into one software package.

[@DataChannel4]

Please also tell me how to calculate indicator reliability step by step in the reflective model? I would really appreciate it if you answer both of my comments. Thank you for your support. Is it just squaring of loadings of each item(factor) in measurement model?

[@ChMahmoodAnwar]

Yes. It's just a square.

[@DataChannel4]

@@ChMahmoodAnwar So I am trying to find the reference for the cut-off value of indicators reliability. Do you have any references for papers or book?

[@ChMahmoodAnwar]

@@DataChannel4 You can use both Hair et al. (2020) and this lecture video as references.

[@DataChannel4]

@@ChMahmoodAnwar Thanks Manoor, But I did not find indicators reliability(for reflective model)>0.50 in Hair Jr, J. F., Howard, M. C., & Nitzl, C. (2020). Assessing measurement model quality in PLS-SEM using confirmatory composite analysis. Journal of Business Research, 109, 101-110...................................can you help where its is? Looks a silly question, but ...... need your help! there is another paper. In Hair et al.2012. which mentions indicators reliability as indicators loading:>0.708 or >0.40.......So this is confusing what is cut-off for indicators reliability actually and where it is ...you may find it for me.

[@ChMahmoodAnwar]

@@DataChannel4 I already told you that I have clarified many points of CCA article in my lecture including indicator reliability value. May be Dr. Hair forgot to mention it in his CCA paper.

[@DataChannel4]

Hi Mansoor, Can you please explain the step-wise procedure of doing predictive validity analysis in smart pls, spss or any other software for the formative model? Do you have any reference paper for this step procedure?

[@ChMahmoodAnwar]

Hi there. I will explain it in a separate video. Meanwhile, you can read this article to test predictive validity of a new construct using correlation & regression analysis www.ekonomie-management.cz/en/archiv/search/detail/1915/

[@DataChannel4]

@@ChMahmoodAnwar Hi Mansoor my model is non-normal so regression seems not a good idea to check predictive validity in SPSS. That would really be helpful if you can make some video or can help to know how I can test predictive validity in this case..

[@ChMahmoodAnwar]

@@DataChannel4 See my latest video on it. Use correlation because it doesn't assume normality.

[@DataChannel4]

@@ChMahmoodAnwar Hi Mansoor, please see your comment above in response to my comment. Here you mentioned, you can use correlation for checking predictive validity if data is not normal. If I use correlation to check predictive validity even if data is normal, will it be correct then? As I have not collected the data for the criterion variable.

[@ChMahmoodAnwar]

@@DataChannel4 Normality of data is another topic. Correlation needs 2 variables, you have not collected criterion, how you will correlate?

[@amaliks13]

I only have formative exogenous, endogenous, and mediator constructs. Should I only use 5 steps for the formative constructs you mentioned?

[@ChMahmoodAnwar]

5 steps are for you. If you have not included independent index in your research previously and already collected data, then you may leave 5th step.

[@DataChannel4]

@@ChMahmoodAnwar On what grounds can we drop predictive validity(5th step), if we have not added an independent index? Can you please share any references? That would be a great help!

[@ChMahmoodAnwar]

@@DataChannel4 Since CCA and application of predictive validity in CCA are relatively new topics, you wouldn't find any citation to this. You can cite this video reference following APA style.

[@DataChannel4]

@@ChMahmoodAnwar Thanks, I will. But I also need to give a logic why I am dropping predictive validity when it is the one step suggested by Hair et al. Can you explain the logic why we are dropping this necessary step?

[@ChMahmoodAnwar]

@@DataChannel4 CCA being a neoteric notion, was not explicitly studied by candidate, therefore, the required data were not collected. Thanks.

[@ragadekar9901]

How to assess both reflective and formative model in one measurement model using CCA

[@ChMahmoodAnwar]

Since PLS is not a SEM technique like CBSEM in AMOS, it is not possible to assess quality of measurement model as we assess in CBSEM. CCA separately assess reflective and formative models' quality.