Table 2. Some corrections to Cadogan and... | Emerald Publishing

Table 2.

Some corrections to Cadogan and Lee (2022)

Questionable statement of Cadogan and Lee (2022)	Proposed correction
“The outputs that PLS produces lie outside the scope of scientific realist inquiry”	Like the ML estimator, PLSc produces consistent estimates for reflective measurement models. Hence, if SEM using ML is suitable for scientific realist inquiry, SEM using PLSc should also be regarded as suitable. Traditional PLS produces estimates for composite models. Scientific realism stays neutral with regard to composition
“PLS … is uniquely designed to construct different stories of the world depending on the research context”	PLS is designed to create composites of observed variables of which the correlation matrix has ‘maximum distance’ to the unit matrix
“[R]ecent work by PLS experts explicitly attempts to link PLS with constructivism.” “… Henseler’s (2017) observation that tools such as PLS are consistent with a constructivist perspective.” “Henseler’s (2017) view of tools like PLS as aligning with constructivism, rather than realism …”	Henseler (2017) suggested that the composite model aligns well with constructivism. Composite models can be estimated using, among others, ML and PLS, which means that ML and PLS are equally strongly linked to constructivism. Henseler (2017) did not view PLS as aligning less with realism
“[T]he numerical results, predictions, and relationships that the PLS method returns are not estimates of real world things, but rather, are explicitly constructed by the analysis method, and hence the analysts themselves”	Every statistical method’s numerical results are produced by the analysis method. PLSc provides consistent parameter estimates for reflective measurement models (Dijkstra and Henseler, 2015a, 2015b), and PLS Mode B provides consistent parameter estimates for composite models (Dijkstra, 2017). Consistency means that estimates converge in probability toward the true value of the parameters given the model is correctly specified
“Further, some PLS advocates agree, identifying PLS as an approach that is entirely constructivist, located in a world where researchers modeling with a ‘composite can be thought of as designers: They design this construct’, explicitly mixing up ‘ingredients … [and arranging them] to form a new entity’ (Henseler, 2017, p. 180), rather than explicitly attempting to measure unobservable variables that ‘exist in nature’ (Henseler, 2017, p. 178). This constructivist interpretation has seriously problematic implications for realists …” (p. 23)	It is questionable whether anybody would agree that PLS is an entirely constructivist approach. Henseler (2017) definitely did not identify it as such. Henseler (2017) recommended specifying reflective measurement models for unobserved conceptual variables that are assumed to exist in nature, and then estimating the reflective measurement model parameters using PLSc. He recommended the composite model as a representation for human-made concepts. Composite models assume “a definitorial relation between a construct and its indicators. This means that the construct is made up of its indicators or elements” (Henseler, 2017, p. 179). In composite models, “the relationships between the indicators and the construct are not cause-effect relationships but rather a prescription of how the ingredients should be arranged to form a new entity” (Henseler, 2017, p. 180). Not all researchers modeling with a composite can be thought of as designers; only those “who introduce a composite can be thought of as designers” (Henseler, 2017, p. 181, emphasis added)
“PLSc’s use of common factors does not elevate PLS to a method that meets the realist’s aspirations for hypothetical causal contact”	Since “causal contact” is a property of a model and not an estimator, and PLSc estimates for reflective measurement models do not differ qualitatively from other estimators for reflective measurement models, SEM with PLSc is a viable methodological option for realists
“[T]he design of PLS means that it will produce different results as a consequence of the varying contexts (e.g., social environments) in which it is undertaken”	The results of PLS remain the same no matter who undertakes the analysis. However, as with maximum likelihood, PLS takes the whole model into account, which means that a construct’s factor loadings can differ if different variables are related to it
“[T]hings that are constructed, like the results from a PLS analysis”	In structural equation modeling, model parameters are estimated, not constructed, no matter which estimator (ML, generalized least squares, unweighted least squares, PLS, etc.) is used
“ Nelson and Stolterman (2012) describe the world in archetypally constructivist terms: for instance they claim that ‘[h]umans did not discover fire – they designed it’ (p. 11), and thus that ‘scientists … can be understood more as design critics than natural scientists’ (p. 27)”	Nelson and Stolterman (2012, p. 11) do not describe the world, but human achievements: “Humans did not discover fire – they designed it. The wheel was not something our ancestors merely stumbled over in a stroke of good luck; it, too, was designed. The habit of labeling significant human achievements as ‘discoveries,’ rather than ‘designs,’ discloses a critical bias in our Western tradition whereby observation dominates imagination.” While constructivists hold that knowledge about the world is constructed (Lyotard, 1984), Nelson and Stolterman (2012) make the point that a large part of today’s world is constructed. At the same time, they do not regard scientific knowledge as being designed: “In the theoretical world of science, we do not think about natural laws or truths as being designed. But, in the real world – the present environment that surrounds all of us – we understand that we ‘create’ as well as ‘discover’ this reality. This is because the real world has many facets of an artificial world and is very much a designed world. In fact scientists have begun to label the present epoch as the Anthropocene era because of the dominant effect human activity has had on global systems, making them ever more unnatural and artificial. Based on this, scientists describing and explaining the world can be understood more as design critics than natural scientists” (p. 27)
“ Henseler (2017) affirms that PLS’s compositions, and the relationships it creates between those compositions, belong in the realm of ‘critical design’, where PLS users essentially ‘imagine that-which-does-not-yet-exist, to make it appear’ Nelson and Stolterman (2012, p. 12)”	Neither Henseler (2017) nor Nelson and Stolterman (2012) use the term ‘critical design,’ so Cadogan and Lee’s claim cannot be affirmed. Nelson and Stolterman (2012) do not say anything about PLS users, but equate design with “the ability to imagine that-which-does-not-yet-exist” (p. 12)

Questionable statement of Cadogan and Lee (2022)	Proposed correction
“The outputs that PLS produces lie outside the scope of scientific realist inquiry”	Like the ML estimator, PLSc produces consistent estimates for reflective measurement models. Hence, if SEM using ML is suitable for scientific realist inquiry, SEM using PLSc should also be regarded as suitable. Traditional PLS produces estimates for composite models. Scientific realism stays neutral with regard to composition
“PLS … is uniquely designed to construct different stories of the world depending on the research context”	PLS is designed to create composites of observed variables of which the correlation matrix has ‘maximum distance’ to the unit matrix
“[R]ecent work by PLS experts explicitly attempts to link PLS with constructivism.” “… Henseler’s (2017) observation that tools such as PLS are consistent with a constructivist perspective.” “Henseler’s (2017) view of tools like PLS as aligning with constructivism, rather than realism …”	Henseler (2017) suggested that the composite model aligns well with constructivism. Composite models can be estimated using, among others, ML and PLS, which means that ML and PLS are equally strongly linked to constructivism. Henseler (2017) did not view PLS as aligning less with realism
“[T]he numerical results, predictions, and relationships that the PLS method returns are not estimates of real world things, but rather, are explicitly constructed by the analysis method, and hence the analysts themselves”	Every statistical method’s numerical results are produced by the analysis method. PLSc provides consistent parameter estimates for reflective measurement models (Dijkstra and Henseler, 2015a, 2015b), and PLS Mode B provides consistent parameter estimates for composite models (Dijkstra, 2017). Consistency means that estimates converge in probability toward the true value of the parameters given the model is correctly specified
“Further, some PLS advocates agree, identifying PLS as an approach that is entirely constructivist, located in a world where researchers modeling with a ‘composite can be thought of as designers: They design this construct’, explicitly mixing up ‘ingredients … [and arranging them] to form a new entity’ (Henseler, 2017, p. 180), rather than explicitly attempting to measure unobservable variables that ‘exist in nature’ (Henseler, 2017, p. 178). This constructivist interpretation has seriously problematic implications for realists …” (p. 23)	It is questionable whether anybody would agree that PLS is an entirely constructivist approach. Henseler (2017) definitely did not identify it as such. Henseler (2017) recommended specifying reflective measurement models for unobserved conceptual variables that are assumed to exist in nature, and then estimating the reflective measurement model parameters using PLSc. He recommended the composite model as a representation for human-made concepts. Composite models assume “a definitorial relation between a construct and its indicators. This means that the construct is made up of its indicators or elements” (Henseler, 2017, p. 179). In composite models, “the relationships between the indicators and the construct are not cause-effect relationships but rather a prescription of how the ingredients should be arranged to form a new entity” (Henseler, 2017, p. 180). Not all researchers modeling with a composite can be thought of as designers; only those “who introduce a composite can be thought of as designers” (Henseler, 2017, p. 181, emphasis added)
“PLSc’s use of common factors does not elevate PLS to a method that meets the realist’s aspirations for hypothetical causal contact”	Since “causal contact” is a property of a model and not an estimator, and PLSc estimates for reflective measurement models do not differ qualitatively from other estimators for reflective measurement models, SEM with PLSc is a viable methodological option for realists
“[T]he design of PLS means that it will produce different results as a consequence of the varying contexts (e.g., social environments) in which it is undertaken”	The results of PLS remain the same no matter who undertakes the analysis. However, as with maximum likelihood, PLS takes the whole model into account, which means that a construct’s factor loadings can differ if different variables are related to it
“[T]hings that are constructed, like the results from a PLS analysis”	In structural equation modeling, model parameters are estimated, not constructed, no matter which estimator (ML, generalized least squares, unweighted least squares, PLS, etc.) is used
“ Nelson and Stolterman (2012) describe the world in archetypally constructivist terms: for instance they claim that ‘[h]umans did not discover fire – they designed it’ (p. 11), and thus that ‘scientists … can be understood more as design critics than natural scientists’ (p. 27)”	Nelson and Stolterman (2012, p. 11) do not describe the world, but human achievements: “Humans did not discover fire – they designed it. The wheel was not something our ancestors merely stumbled over in a stroke of good luck; it, too, was designed. The habit of labeling significant human achievements as ‘discoveries,’ rather than ‘designs,’ discloses a critical bias in our Western tradition whereby observation dominates imagination.” While constructivists hold that knowledge about the world is constructed (Lyotard, 1984), Nelson and Stolterman (2012) make the point that a large part of today’s world is constructed. At the same time, they do not regard scientific knowledge as being designed: “In the theoretical world of science, we do not think about natural laws or truths as being designed. But, in the real world – the present environment that surrounds all of us – we understand that we ‘create’ as well as ‘discover’ this reality. This is because the real world has many facets of an artificial world and is very much a designed world. In fact scientists have begun to label the present epoch as the Anthropocene era because of the dominant effect human activity has had on global systems, making them ever more unnatural and artificial. Based on this, scientists describing and explaining the world can be understood more as design critics than natural scientists” (p. 27)
“ Henseler (2017) affirms that PLS’s compositions, and the relationships it creates between those compositions, belong in the realm of ‘critical design’, where PLS users essentially ‘imagine that-which-does-not-yet-exist, to make it appear’ Nelson and Stolterman (2012, p. 12)”	Neither Henseler (2017) nor Nelson and Stolterman (2012) use the term ‘critical design,’ so Cadogan and Lee’s claim cannot be affirmed. Nelson and Stolterman (2012) do not say anything about PLS users, but equate design with “the ability to imagine that-which-does-not-yet-exist” (p. 12)