Methods of Reporting Research Results of Second-order Construct of PLS-SEM

Abstract

Studies involving the reporting of research results of the second-order construct of PLS-SEM are distinct from studies in which there are only first construct models. Furthermore, they also provide distinct results when the studies are of different types and use different approaches. The literature mostly focuses on how to construct and assess higher-order models and lacks to provide detail concerning the methodology of reporting, particularly as it pertains to composite models. Thus, this study aims to help readers to understand the methodology of reporting the results of hierarchical component models. The study used data on service quality, trust, and loyalty of E-commerce in Thailand. This study used two models, (1) the reflective-reflective type and the repeated indicator approach in which the algorithm was Mode A consistent or consistent partial least squares and factor. (2) Used the reflective-formative type and a two-stage approach in which the algorithm was composite and Mode B. The results of the structure model for both models were not very different. The repeated indicator approach had no discriminant validity on the measurement model thus ADANCO could not generate a global model fit. On the other hand, the two-stage approach evoked discriminant validity thus achieving adequate parameters for a model fit, the measurement model and a structure model.