SMAP - Mediation vs. Moderation

Steps

1. Is my research question about mediation or moderation?

   Mediation and moderation analysis are sometimes confused, but they shouldn’t, because conceptually they are very different. Both mediation and moderation describe a relationship between at least 3 variables: at least 2 independent variables (IVs), one of which is the moderator or mediator, and 1 dependent variable (DV).

   

   In the figures above, the left figure depicts mediation, and the right figure depicts moderation.

   If your research question is about mediation, you want to investigate whether the effect of an independent variable X on a dependent variable Y runs via a mediator. That means that the effect X on Y can in fact be ascribed to the mediator (to some degree), or put differently: X has an effect on the mediator, and the mediator in turn has an effect on Y. For example: a letter sound training (X) increases children’s phonological awareness (Med), which leads to better reading skills (Y).

   If your research question is about moderation, you want to investigate whether the effect of an independent variable X on a dependent variable Y is dependent on a third variable (the moderator). That is, whether a moderating variable has an effect on (the strength of) the relationship between X and Y. In statistical terms, this means that there is an interaction between two IVs on the DV.

   Interaction effects occur in different patterns. One possibility is that the effect of X on Y changes its direction depending on the value of the moderator: the effect of X on Y is positive at some levels of the moderator, and negative at other levels of the moderator. For example: age (Mod) moderates the relationship between music volume at a festival (X) and enjoyment (Y): this relationship is positive for young people but negative for old people.

   However, another possible pattern is that the effect of X on Y differs in strength depending on the value of the moderator: the effect of X on Y then has the same direction for all values of the moderator, but the effect of X on Y is stronger at some levels of the moderator than at other levels.

   For example, offering a monetary reward to do well on an IQ-test (X) has a positive effect on IQ scores (Y), but the strength of this effect is moderated by motivation (Mod). The effect is stronger for people with low motivation and weaker for people with high motivation.