Understanding PLS path modeling parameters estimates: a study based on Monte Carlo simulation and customer satisfaction surveys

PLS path modeling (Partial Least Squares path modeling) has found increased interests since being used in the context of marketing studies. In this paper, we use Monte Carlo simulation and real life datasets to investigate the effects of changes in the model specification or in the data. We first introduce PLS path modeling and some already well-known properties of this approach; we then conduct studies to estimate robustness of the approach to sample size, blocks size, distribution of the number of manifest variables inside the blocks and modification of the manifest variables values. All these studies are carried out using Monte Carlo simulations and real datasets from customer satisfaction surveys in the French power supply sector. We conclude with some general remarks on a framework to better analyse data and models using PLS path modeling.