Quantifying the Impact of Measurement Errors in Consistent Linear Partial Least Squares Structural Equation Modelling: A Monte Carlo Investigation

Abstract

Measurement errors wield the potential to introduce uncertainties and inaccuracies, casting shadows on data quality and jeopardizing the integrity of structural relationships. Notably robust against measurement errors, Partial Least Squares Structural Equation Modelling (PLS-SEM) has historically maintained a reputation for resilience. However, recent insights have unveiled its susceptibility to these errors, instigating a revaluation of its standing in the Structural Equation Modelling landscape. Overlooking measurement errors in PLS-SEM carry consequential repercussions, notably tainting the accuracy of structural relationships and introducing bias. This effect becomes particularly pronounced when dealing with an insufficient understanding of the intricate structural dynamics. Unfortunately, PLS-SEM currently lacks an all-encompassing remedy to address this concern. Consequently, the quantification of measurement errors impact in PLS-SEM gains paramount importance, fostering a demand for innovative strategies to propel its effectiveness forward. Notably, contemporary investigations have unmasked PLS-SEM's vulnerability to non-orthogonal errors. This revelation challenges the notion of its imperviousness to the detrimental influence of measurement errors, necessitating a comprehensive evaluation of its performance under such conditions. This study leveraged simulated data to extract empirical findings and employed parameters biasedness analysis. This analysis led to the determination that the stability of the PLS-SEM algorithm is compromised when exposed to diverse measurement error scenarios. Consequently, the outcomes generated exhibit both instability and bias. This bias becomes increasingly conspicuous as the magnitude of measurement errors intensifies. Thus, the study proposes avenues for elevating the robustness of PLS-SEM.