Virtual useR! 2024

Structural equation modeling (SEM) is often used to test theories, verify model measurement properties, and obtain unbiased estimates, a widespread modeling approach for composite hypotheses. However, power analysis is seldom considered in SEM studies to ensure the required sample size needed to achieve adequate power for detecting the hypothesized effect. This is due partly to the lack of a comprehensive SEM power analysis software package. The session introduces the intelligible semPower package that provides both global and local power analysis for establishing the power of a model and the specific hypothesis respectively. Three models will be illustrated. First, by varying the factor loadings of a confirmatory factor analysis model to determine power and sample size. Second, varying two loadings concurrently of a mediation model to determine the required sample size, and varying the covariance of a latent growth model to test for a local power analysis. The R syntax is illustrated to show the usefulness of using the package semPower, and graphing the model via the package semPlot. Users will find the syntax simple and can easily carry out power analysis for their studies.