An evaluation of PLS based complex models: the roles of power analysis, predictive relevance and GoF index

Structural equation modeling (SEM) is an important tool to estimate a network of causal relationships linking two or more complex concepts. The PLS approach to SEM, also known as component based SEM, is becoming more prominent for estimating large complex models due to its soft modeling assumptions. This ‘soft modeling’ refers to the greater flexibility of PLS technique in developing and validating the complex models. However, to establish rigor in such complex modeling, this study highlights the critical roles of power analysis, predictive relevance and GoF index. The findings of the study show that power analysis is essential to establish conjectures based on IT artifacts, predictive relevance is vital to measure how well observed values are reproduced by the model and finally, GoF index is crucial for assessing the global validity of a complex model.