Development of a polar extreme method for use in partial least squares SEM

The use of polar extremes as a data segmentation approach is a common practice to maximize variance within multivariate analysis. In this research, we propose the first application of the polar extremes methodology within partial least squares structural equation modeling (PLS-SEM). Discussed is the impact of unequal weights of indicators among latent constructs within PLS-SEM and the importance of this methodology to the current established PLS-SEM analysis approach. Using a case study approach, we demonstrate the method. Developing two different models within the case allows comparison of the path coefficients and percentage of variance explained between the current PLS-SEM application and the proposed polar extremes methodology.