ABSTRACT
An excellent computational tool for observed variable path analysis-based moderation and mediation is PROCESS. Moderated multiple regression (MMR) is an inferential approach to examining moderation that consists of comparing two least-squares regression equations. The usual assumptions of multiple linear regressions are applicable for moderated multiple regressions and include: linearity; residuals that are homoscedastic, normally distributed, and independent; and lack of multicollinearity. Estimating sample size for mediation models is more complicated than with multiple linear regressions. The ratio of the indirect effect to total effect is an effect size for mediation models that is sometimes reported. A model estimated without mean centering is mathematically equivalent to a model estimated with mean centering. Low reliability of modeled variables is so problematic that measurement error is considered by some researchers to be the most impactful factor on power in MMR.
