Moderation
Also known as: Moderator, Interaction effect, Effect modification
When the strength or direction of an effect depends on a third variable.
What it means
Moderation occurs when the effect of one variable on an outcome changes across levels of another variable, the moderator, so the relationship is conditional rather than uniform. Statistically it is tested as an interaction term, and it answers the question of for whom, or under what conditions, an effect holds. Moderation is conceptually distinct from mediation, because a moderator specifies when or how strongly an effect operates, whereas a mediator explains the mechanism by which it operates. Detecting interactions reliably typically demands substantially larger samples than detecting main effects, a fact that makes many published moderation claims fragile.
Examples
A drug lowers blood pressure in older patients but not younger ones, so age moderates the drug's effect.
A reminder email lifts attendance among people who already meant to come but does nothing for the uninterested, so prior intention moderates whether the reminder works at all.
Free tutoring raises grades sharply in schools with small classes and barely moves them in crowded ones — class size decides how much the same programme delivers.
First described in Baron & Kenny (1986); long regression tradition.