Andrew Conway is a Psychology Professor in the Division of Behavioral and Organizational Sciences at Claremont Graduate University in Claremont, California. He has been teaching introduction to statistics for undergraduate students and advanced statistics for graduate students for 20 years, at a variety of institutions, including the University of South Carolina, the University of Illinois in Chicago, and Princeton University.
Moderation and mediation sound alike, but in reality they are quite different. This course will get you acquainted with these complex analytical techniques based on multiple regression. Special attention will go to centering predictors as a way to improve interpretability of results.
In this chapter professor Conway will give an introduction to the concepts behind moderation while you will walk through an example. You will define models with and without moderation, and study the difference in performance between these. Furthermore, it will be illustrated how you can visualize the effects of moderation.
Centering predictors means to take your predictor variable and to put it in deviation form. Centering predictors makes the interpretation of a moderation analysis much simpler, but unfortunately it often is a tedious work. The chapter will focus on the conceptual part behind centering, as well as doing the calculations in R.
Just like moderation, mediation is a multivariate approach as well. The mediator variable is designed to account for, or to explain the relationship between a predictor and an outcome. You will learn how to do a mediation analysis via a series of multiple regressions, which is the standard approach. The techniques to conduct a mediation analysis in R are illustrated as well.