Title: Statistical power with respect to true sample and true population paths: a PLS-based SEM illustration

Authors: Ned Kock; Murad Moqbel

Addresses: Division of International Business and Technology Studies, Texas A&M International University, Laredo, TX 78041, USA ' Health Information Management Department, University of Kansas Medical Center, Kansas City, KS 66160, USA

Abstract: Monte Carlo experiments aimed at assessing the statistical power of structural equation modelling (SEM) techniques typically focus on true population path coefficients, ignoring true sample path coefficients. We demonstrate the limitations stemming from such practice in statistical power assessments. This is done in the context of SEM techniques employing the partial least squares (PLS) method, where power claims have led to much recent debate. We show that the sample sizes at which power is greater than .8 differ significantly when we consider true population and true sample paths, and that the difference increases with decreases in the magnitudes of the paths being considered. Finally, we illustrate empirically how these differences affect the conclusions we can draw from the analysis of a relatively small sample of size 193.

Keywords: information systems; statistical power; structural equation modelling; SEM; latent variables; partial least squares; PLM; Monte Carlo simulation; true sample path; true population path.

DOI: 10.1504/IJDATS.2016.081365

International Journal of Data Analysis Techniques and Strategies, 2016 Vol.8 No.4, pp.316 - 331

Received: 24 Jan 2015
Accepted: 21 Jul 2015
Published online: 06 Jan 2017 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article