Moderation Analysis Calculator

Moderation analysis tests whether the relationship between an independent variable (X) and a dependent variable (Y) changes depending on the level of a third variable — the moderator (W). If moderation is present, the interaction term X×W will be statistically significant, meaning the effect of X on Y differs across values of W.

Moderation asks 'when' or 'for whom' an effect occurs — the moderator W changes the strength or direction of X→Y. Mediation asks 'how' or 'why' an effect occurs — the mediator M transmits the effect of X to Y. Both involve a third variable but serve fundamentally different explanatory roles.

You need the regression coefficients (b1 for X, b2 for W, b3 for the X×W interaction), their corresponding standard errors, your sample size N, total number of predictors, and the R² values for both the base model (without interaction) and the full model (with interaction).

The interaction coefficient b3 represents how much the effect of X on Y changes for each one-unit increase in W. A significant b3 (p < your chosen α) indicates that W moderates the X→Y relationship. A positive b3 means the effect of X strengthens as W increases; a negative b3 means it weakens.

ΔR² (delta R-squared) is the additional variance in Y explained by adding the interaction term X×W to the base model. A meaningful ΔR² alongside a significant F-change confirms that the moderator genuinely improves model fit beyond the main effects alone.

The t-statistic for b3 is calculated as b3 divided by its standard error (SE). It quantifies how many standard errors the interaction coefficient is away from zero. The larger the absolute t-value, the more evidence there is against the null hypothesis that no moderation exists.

Moderation analysis generally requires larger samples than simple regression because interaction effects tend to have smaller effect sizes. As a rule of thumb, a minimum of 50–100 participants is recommended, though power analyses often suggest 200+ for detecting small to medium interaction effects reliably.

Standardizing (mean-centering) X and W before computing the interaction term X×W is strongly recommended. It reduces multicollinearity between the main effects and the interaction term, making the coefficients for b1 and b2 more interpretable as effects at the mean of the other variable.