Estimation issues with PLS and CBSEM: Where the bias lies!

Abstract

Discussions concerning different structural equation modeling methods draw on an increasing array of concepts and related terminology. As a consequence, misconceptions about the meaning of terms such as reflective measurement and common factor models as well as formative measurement and composite models have emerged. By distinguishing conceptual variables and their measurement model operationalization from the estimation perspective, we disentangle the confusion between the terminologies and develop a unifying framework. Results from a simulation study substantiate our conceptual considerations, highlighting the biases that occur when using (1) composite-based partial least squares path modeling to estimate common factor models, and (2) common factor-based covariance-based structural equation modeling to estimate composite models. The results show that the use of PLS is preferable, particularly when it is unknown whether the data's nature is common factor-or composite-based.

The Impact of Managerial and Adaptive Capabilities to Stimulate Organizational Innovation in SMEs: A Complementary PLS–SEM Approach

“…More specifically, SmartPLS 3 [116,117] was used to estimate the research model (for detailed reasons of why and when to use PLS-SEM, see for example, Richter, Cepeda, Roldán, & Ringle [118]). Despite a surprising level of animosity towards PLS-SEM [118][119][120][121], PLS-SEM has been widely accepted by the scholarly community, including authors, reviewers, and editors [122][123][124][125]. The following points summarize why this study adopted PLS-SEM instead of Linear structural relations (LISREL) or Analysis of a moment structures (AMOS) as more suitable statistical techniques: (1) the structural model is complex, and contains four series of dependent relationships [118,126]; (2) the research objective of the structural model is prediction oriented, and explaining the variance in key target constructs [126,127]; (3) this study analyzes the relationships between managerial capability, adaptive capability, and organizational innovation; which is being considered in the initial stages of theory development, therefore motivated us to investigate the related phenomena in this emerging area [118]; (4) the sample size (n = 210) is also believed to be relatively small [127], finally; (5) this study also takes advantage of PLS-SEM in terms of its less rigorous requirements for restrictive assumptions, which motivates researchers to develop and estimate such models through enabling them to avoid additional limiting constraints [117,128].…”

The Effect of Absorptive Capacity on Green Customer Capital under an Organizational Unlearning Context

“…First, we had a composite measurement model. Both theoretical studies [97][98][99] and empirical simulation studies [100,101] recommend and support the use of PLS for composite models. Second, as per Chin's [102] indications, we used PLS because we employed latent variable scores in subsequent analysis for modeling a second-order multidimensional construct, applying the higher-order component two-stage approach [103].…”

Examining the Role of Childhood Experiences in Developing Altruistic and Knowledge Sharing Behaviors among Children in Their Later Life: A Partial Least Squares (PLS) Path Modeling Approach

“…The knowledge donating and knowledge collecting scales have been developed by van den Hooff and Hendrix [77], both are first-order reflective constructs contain four items. Following the recent recommendations in Henseler [73], Rigdon, Sarstedt, and Ringle [78], Sarstedt, Hair, Ringle, Thiele, and Gudergan [79], and van Riel, Henseler, Kemény, and Sasovova [80], all constructs are estimated in Mode A, at the item, the first-order and the second-order construct level. Finally, this study follows Wright, Campbell, Thatcher, and Roberts [81] to use the two-stage approach to evaluate the all hierarchical second-order constructs [82].…”