forum.smartpls.com

Hi,
I am unsure about modeling, estimating, and testing a third-order construct made up of formatively measured first- and second-order constructs with SmartPLS.
The most current guidelines by Van Riel et al. (2017) do not cover this type of hierarchical constructs yet and in Ringle et al. (2012) the repeated indicators approach or even a mixture of the repeated indicators approach and the use of latent variable scores in a two-stage approach is recommended (for second-order constructs!).

Actually, I intented to follow the approach by James Gaskin presented in the video "SmartPLS 3 2nd and 3rd order factors using the repeated indicator approach" (https://youtu.be/LRND-H-hQQw). But when I noticed a caution by Jörg Henseler not to follow this approach, I became uncertain about it. However, in his comment on this video he argues that the approach yields inconsistent estimates if the first-order constructs are reflective. But in my case, they are NOT (but formative)! So, the approach might work for me?
My model comprises 1 formative HOC ("Kohärenz") on 3rd order, 2 formative HOCs (V.K. & H.K.) on 2nd order, with 3 formative LOCs each (FW, FD, BW and FW-FD, FW-BW, FD-BW) on 1st order with 2-3 formative indicators (see PLS screenshot attached, already with repeated indicators approach).

I would highly appreciate any support about this question!

Kind regards,
Jan
-----------------------------------------
Ringle, C.M., Sarstedt, M. and Straub, D.W. (2012), “Editor’s comments: a critical look at the use of PLS-SEM in MIS quarterly”, MIS Quarterly, Vol. 36 No. 1, pp. iii-xiv.
​
van Riel, Allard C. R. van; Henseler, Jörg; Kemény, Ildikó; Sasovova, Zuzana (2017). Estimating hierarchical constructs using consistent partial least squares: The case of second-order composites of common factors. Industrial Management & Data Systems, 117 (3), 459-477.​​​

You are correct, especially in your situation the concerns regarding consistency of the estimates within the higher-order construct are not a problem as you have no reflective (factor model type) indicators in your construct. Thus, you can use the two-stage approach (which would basically mean to use only the second-order as indicators on the third-order construct) or the repeated-indicator approach.
With the repeated indicator approach the guidelines from Becker et al. 2012 should also apply to use path weighting scheme for the inner model estimation and formative indicator direction (Mode B) on the higher-order constructs.

Becker, J. M., Klein, K., & Wetzels, M. (2012). Hierarchical latent variable models in PLS-SEM: guidelines for using reflective-formative type models. Long Range Planning, 45(5), 359-394.

Thanks for the immediate reply!
Choosing between the two-stage and the repeated-indicator approach we would prefer the two-stage as it also allows assessing the HOC(s) using CTA-PLS, which is not feasible with the repeated-indicator approach (see Hair et al., 2018), correct?
However, with the two-stage-approach (which in our case would be a "three-stage-approach") we still wonder how to best begin in the first stage, that is, estimating first-order construct scores based on the model without the second-order construct(s): by removing the two 2nd order HOCs from the model the six LOCs on 1st order remain isolated and SmartPLS won`t run. So how does one estimate the LOC scores? Is it necessary to begin with a full model or is there any alternative way to achieve model identification (e.g. adding an exogenous variable)?
thanks in advance!
kind regards,
Jan

-----------------------------------------
Hair, J. F., Sarstedt, M., Ringle, C. M., & Gudergan S. P. (2018). Advanced Issues in Partial Least Squares Structural Equation Modeling (PLS-SEM), Sage: Thousand Oaks.

Hi, I have similar problem for my Phd Thesis. I want to develop 3rd order (Reflective for 1st order, Formative for 2nd and 3rd order). So, may I know which method is best to evaluate it?

Appreciate your help

I have the same issue (first order reflecteive 2nd and 3rd are formative). @ fhm5020 have you found a solution?