ABSTRACT
This chapter describes the analysis of moderator effects, followed by data examples. It presents mediation effects with emphasis on the importance of the assumptions required to claim mediation, and provides illustrative data examples. A moderator is a variable that affects the strength of the relationship between the main predictor of interest and the outcome. The moderator can be categorical (qualitative) or dimensional (quantitative). Moderator effects need to be interpreted only with respect to the chosen metric and in the context of the available sample size. The chapter considers the situation when the outcome is not normally distributed and hence a non-linear link function is used to relate the outcome to the predictors. In observational studies, assessment of moderator effects is more complicated, since exposure may be related to the potential moderators. The chapter concludes with a summary and reiteration of the challenges in moderation and mediation analyses.
