CFA with PLS

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
Post Reply
phadaya
PLS Junior User
Posts: 2
Joined: Mon Dec 05, 2011 1:28 pm
Real name and title:

CFA with PLS

Post by phadaya »

what is the procedure to conduct a CFA with smartpls? Let say I have three constructs, can I conduct a CFA without conecting the three contructs to each other?
Thx.
Pierre
User avatar
Hengkov
PLS Super-Expert
Posts: 1599
Joined: Sun Apr 24, 2011 10:13 am
Real name and title: Hengky Latan
Location: AMQ, Indonesia
Contact:

Post by Hengkov »

Hi Pierre,
1. Running bootstrap procedure for test CFA.
2. You can connecting or no connecting construct for test CFA, same result.
Regards,
Hengky
Gammadrac
PLS Junior User
Posts: 8
Joined: Mon Feb 06, 2012 10:19 am
Real name and title:

Post by Gammadrac »

I have the same problem and need to perform a CFA with Smartpls
But If you not connect the constructs your Smartpls says: "There must be a vallid path for your calculations" and constructs become red,

For any suggestion also in general to perform CFA in Smartpls I'm very thankful! Also thx for developing Smartpls and providing this support to others :-)
christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl »

Hey,
One way to do a CFA directly in PLS is to use two constructs. For both constructs you use the same indicators. For one construct you use the items reflectively and for the other one formatively. Then you connect the formatively measured construct with the reflectively measured construct. Whereby the path direction goes from the formative measured construct to the reflective measured construct.
This is just an idea…
Christian
Gammadrac
PLS Junior User
Posts: 8
Joined: Mon Feb 06, 2012 10:19 am
Real name and title:

Post by Gammadrac »

Hm actually I want to measure 10 reflective constructs. I would do it normaly with AMOS but AMOS is not suited for my sample size and produces very bad results. Hoped that there would be a way to use PLS as I also use it for modeling.

E.g.
Promotional Strategies for Small Businesses: Group Buying Deals
Tracy L. Tuten, Christy Ashley
The PLS model was analyzed using SmartPLS version 2.0 M3. First, the measurement model was tested to ensure the measures used to capture the constructs are valid and reliable
These authors did it somehow with smartpls by first estabslishing a measurement model without any connections as it seems and later normally drawing connections and testing their structural model.... But question is how to do a measurement model and thus a CFA. If you have no connection the software does not run as far as it seems.
For any hint, thanks a lot :-)
christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl »

Hey Martin,

In my opinion the quote is a little bit confusing. The evaluation in a PLS model is step wise that doesn’t mean that there is a separate estimation before you test your model. You always test your model as a whole. But in the first step you only consider the measurement models. After good critical values for your measurements you can evaluate the path model in the second step. This is the normal way. Please check following working paper (only German) for how to perform the measurement and path model evaluation:

https://www.unibw.de/ifc/team/wimi/nitz ... load/down1

Hope that helps!
Christian
Gammadrac
PLS Junior User
Posts: 8
Joined: Mon Feb 06, 2012 10:19 am
Real name and title:

Post by Gammadrac »

It does. Thanks a lot Christian :-)
karenkim
PLS Junior User
Posts: 6
Joined: Sat Jan 07, 2012 5:06 pm
Real name and title:

English version?

Post by karenkim »

It looks useful.
Is there any English version for the measurement model in PLS?
christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl »

If you search after nice explanations how to perform the evaluation of measurement in a PLS model then you can find some good hints here:

http://php.portals.mbs.ac.uk/Portals/49 ... cs-PLS.pdf

Greetings,
Christian
imran77
PLS Junior User
Posts: 4
Joined: Tue Mar 20, 2012 4:17 am
Real name and title:

Post by imran77 »

how did you conduct a CFA without connecting to other constructs?
IMRAN
User avatar
Diogenes
PLS Super-Expert
Posts: 899
Joined: Sat Oct 15, 2005 5:13 pm
Real name and title:
Location: São Paulo - BRAZIL
Contact:

Post by Diogenes »

Hi,

Even in the AMOS (or LISREL, EQS… covariance based), we connect all LV when running a CFA (double arrows), meaning that the LV correlation could be any value [1].

In the PLS-PM the procedure is the same:
1) Draw the LV with their indicators
2) Connect all LV (single arrows and it doesn’t matter the direction), one with another (without feedbacks, nonrecursive model)
3) Run the PLS algorithm with the “factor” weighting scheme. (The relations between LV will be computed as correlations) [2]
4) Do not use the results that are included in the picture of the model (structural relations)
5) Use the results that are in the “Report”: LV correlation (*); Overview; cross loadings, to assess the measurement model.


(*) If you want, you could redraw the model, putting the correlations in double arrows (as it is done in LISREL). The XLSTAT-PLSPM software already has this option.

[1] BROWN, T. A. Confirmatory Factor Analysis for Applied Research. New York: The Guilford Press, 2006.

[2] TENENHAUS, M.; ESPOSITO VINZI, V.; CHATELIN, Y.-M.; LAURO, C. PLS path modeling. Computational Statistics & Data Analysis, v. 48, n. 1, p. 159-205, 2005.


Best regards,

Bido
User avatar
Diogenes
PLS Super-Expert
Posts: 899
Joined: Sat Oct 15, 2005 5:13 pm
Real name and title:
Location: São Paulo - BRAZIL
Contact:

Re: CFA with PLS

Post by Diogenes »

CFA in SmartPLS to comput scores:

[1] If you had a structural model, and the formative LVs were in exogenous position, we could run the model with covariance based on lavaan (a R package - http://lavaan.ugent.be/tutorial/index.html). But, if the purpose is to compute the factor scores, covariance based is not the best choice (see DiStefano et al., 2009).
DiStefano, C., Zhu, M., & Mîndrilă, D. (2009). Understanding and Using Factor Scores:Considerations for the Applied Researcher. Practical Assessment, Research & Evaluation, 14(20), 1–11. Retrieved from http://pareonline.net/getvn.asp?v=14&n=20


[2] Reference about factor weighting scheme as PCA: Lohmoller (1989, p.42):
Lohmöller, J.-B. (1989). Latent variable path modeling with partial least squares. Heidelberger: Physica-Verlag.


[3] One example is Asyraf and Afthanorhan (2013), but they did not include all relations in the PLS model (p.201).
Asyraf, W. M., & Afthanorhan, B. W. (2013). A comparison of partial least square structural equation modeling (PLS-SEM) and covariance based structural equation modeling (CB-SEM) for confirmatory factor analysis. International Journal of Engineering Science and Innovative Technology (IJESIT), 2(5), 198–205.


[4] A second example: Gefen and Straub. (2005).
Gefen, D., & Straub, D. (2005). Pls-Graph: Tutorial and Annotated Example. Communications of the Association for Information Systems, 16, 91–109.

Best regards,

Bido
Post Reply