Abstract & Overview
The use of the partial least squares (PLS) approach for structural equation modeling (SEM) has been experiencing explosive growth, particularly in the last few years. The calculation of p-values is extensively used for hypothesis testing in PLS-SEM. Such calculation typically relies on standard errors estimated via bootstrapping. This leads to unstable p-values and prohibitive computational demands when very large samples are analyzed. We discuss two calculation methods relying on exponential adjustment that generate stable standard errors and p-values,and that have minimal computational requirements. A Monte Carlo experiment shows that the methods yield estimates of the actual standard errors that are generally consistent with bootstrapping, and often more precise. The methods are implemented as part of the software WarpPLS, starting in version 5.0.