Month: Jun 2019

The Public Administration Manifesto II

stephensphdbloggLeave a comment

Ling Zhu, Christopher Witko, Kenneth J Meier, The Public Administration Manifesto II: Matching Methods to Theory and Substance, Journal of Public Administration Research and Theory, Volume 29, Issue 2, April 2019, Pages 287–298, https://doi.org/10.1093/jopart/muy079


I read this article with anticipation and was not let down. The paper is the result of a “methods symposium” “that will appear in this and the next two issues of the
Journal of Public Administration Research and Theory”. It links the
articles in this symposium together into a combination of a descriptive
“state of the art” and a normative impetus on forking out a path
forward. “

The article gave me the “big picture” on use of methods in PA-research – and I will surely return to it as a guide showing me where to go for digging further into methodological questions. If you want a up-to-date vitamin-injection concerning methodological challenges in PA research: look no further.

Four conclusions:
1 Self-reflective use of methods is essential.
2 Methodological pluralism is necessary – challenging the division of
qualitative and quantitative approaches
3 Generalizability and replicability are real and vital challenges
4 PA needs a stronger arena for “methodological debates regarding best
practices and sophisticated methods in different substantive research
areas.”

I will hunt down the main articles referred to in this article, and they will appear in this space in the coming months.

 

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