9. Causal Steps to Establish Mediation: Steps 3 & 4

Causal Steps 3 and 4 in a mediation analysis

Steps 3 and 4 are conducted simultaneously using multiple regression. Step 3 consists of regressing the outcome variable y onto both the mediator, m, and the predictor, x to provide an estimate of path b.

Note: it is not sufficient just to correlate the mediator with the outcome; the mediator and the outcome may be correlated because they are both caused by the initial variable X. Thus, the initial variable must be controlled in establishing the effect of the mediator on the outcome.

Step 4 is simply the estimate c' of the effect of x on y, controlling for the effect of m, which is output in the same multiple regression. If c' is zero, complete mediation of the effect has occurred. If c' is not zero, then partial mediation has occurred.

In our example we use multiple regression to regress group attraction on to both visual anonymity and self-categorization (Steps 3 and 4). From this we obtain the estimate of the effect of Self-categorization on group attraction, controlling for visual anonymity (path b), and the effect of visual anonymity on group attraction, controlling for Self-categorization (path c').

If the estimate of path c' is not zero, then the effect of visual anonymity on group attraction is only partially mediated by self-categorization.

Group attraction is regressed on to Self-categorization and visual anonymity

NEXT: 10. Barron and Kenny's (1986) Criteria for Mediation


Statistics Training: Introduction to Path Analysis


If you found this article useful, please share so others can find it.