PLS-SEM

Consistent PLS-SEM – PLSc

stephensphdbloggLeave a comment

Article: Dijkstra, Theo & Henseler, Jörg. (2015). Consistent Partial Least Squares Path Modeling. MIS Quarterly. 39. 10.25300/MISQ/2015/39.2.02.

link: https://misq.org/catalog/product/view/id/1701

The PLS-SEM tool got a valuable update/option with “Consistent PLS-SEM” = PLSc. No single approach is “the one and only” tool. The article explains when you should know about some of the shortcomings of classical PLS-SEM, and demonstrates some of the advantages to the updated PLS algorith called “PLSc”. Especially “PLSc provides a correction for estimates when PLS is applied to reflective constructs: The path coefficients, inter-construct correlations, and indicator loadings become consistent” (Dijkstra et al. 2015).
I found the article to be both important and clear – supplying good reason to be aware of which situations classical PLS-SEM has shortcomings. The program SMART-PLS lets you run the same data with either the classical allgorithm or the newer PLSc – but you should know why you choose either or. This article helps you know.
I ran my data with both classical PLS-SEM and PLSc. The difference in results were noteworthy, both in the structural model and the measurement model. I both had data with non-normal distribution and I used reflective measurement models – so PLSc seemed to be the best choice of the two.
The worst of the math is put in the appendix – leaving the article itself unriddled.

More related to this: Another article by the same author deals with PLSc in nonlinear SEM’s: Dijkstra, T. K., & Schermelleh-engel, K. (2014). Consistent partial least squares for nonlinear structural equation models. Psychometrika, 79(4), 585-604. doi:http://dx.doi.org.ezproxy.hioa.no/10.1007/s11336-013-9370-0


PLS-SEM for ‘dummies’

stephensphdbloggLeave a comment

Article: Haenlein, Michael, and Andreas M. Kaplan. 2004. “A Beginner’s Guide to Partial Least Squares Analysis.”  Understanding Statistics 3 (4):283-297. doi: 10.1207/s15328031us0304_4.

Full article: here

PLS-SEM is a powerful statistical method to analyze how several variables, each consisting of a number of indicators, influence eachother. The approach is referred to as a “silver bullet” – a “quasi-standard in marketing and management research when it comes to analyzing the cause-effect relations between latent constructs” (Hair et al. 2004)

Although beeing from 2004, and PLS-SEM has developed since then, this article gives a reasonable and clear description of this statistical tool.
The article compares PLS-SEM to other approaches, especially CB-SEM. CB-SEM is similar, but also quite different, being covariance-based, whereas PLS-SEM is variance-based.
It’s quite easy to use a PLS-SEM analysis tool, like Smart-PLS, but if you havent used second-generation algorithms like PLS-SEM, its a bit like getting into a rocket-ship, after being used to riding a scooter (ok, slightly exaggerated). To know what the results actually mean, you ought to have at least a basic understanding of whats under the hood. This article lets you have a peek.

Of course – if you want to dig in – have a go at this : https://www.amazon.com/Partial-Squares-Structural-Equation-Modeling/dp/148337744X

Reporting well when using PLS-SEM

stephensphdbloggLeave a comment

Article: Joseph F., Jr, Hair, & Risher, Jeff & Sarstedt, Marko & Ringle, Christian. (2018). When to use and how to report the results of PLS-SEM. European Business Review. 31. 10.1108/EBR-11-2018-0203.
Link: https://www.emeraldinsight.com/doi/abs/10.1108/EBR-11-2018-0203

SEM-analysis, Structural Equation Modelling, is described as a second-generation technique within multivariate methods (Hair et al. 2017). First-generation methods comprise of exploratory methods like cluster analysis, exploratory factor analysis and multidimensional scaling, and confirmatory methods like analysis of variance, logistic regression, multiple regression, and confirmatory factor analysis. Second generation techniques go further in statistical strength and flexibility with the use of Partial Least Squares Structural Equation Modelling (PLS-SEM) as an exploratory tool and Covariance-Based Structural Equation Modelling (CB-SEM) as a confirmatory tool.

Hair (2017) describes some of the benefits of using SEM-analysis approach as

  • the ability to use composite variables (often called variates),
  • the ability to use unobservable (latent) variables – through indicator variables
  • the ability to account for measurement error in observed variables
  • simultaneously examine relationships among measured variables and latent variables.

SEM-analysis is used as a powerful predictive statistical approach – the analysis of the sample giving predictability to the population.

This article shows the most recent update on how to report the results when using PLS-SEM. I use SMART-PLS 3 as my preferred program for doing PLS-SEM analysis, and it gives me a lot of the data which Hair et.al proscribes in this article. Older articles do not neccesarily live up to the most recent standards of reporting – therefore this article is a good help to reporting well.