The mediated regression approach follows the
guidelines as outlined by Baron & Kenney (1986). Their analyses require three
separate equations to be estimated. The first equation involves regressing the
mediator variable on the predictor variable. The second equation entails
regressing the criterion variable on the predictor variable. Finally, the third
equation involves regressing the criterion variable concurrently onto the
predictor variable and the mediator variable.
Furthermore, Baron and Kenny (1986) outlined
four conditions that must be met:
- Condition 1. The
antecedent/predictor variable must be significantly related to the
mediator
- Condition 2. The
antecedent/predictor variable must be significantly related to the
criterion variable
- Condition 3. The mediator
variable must be significantly related to the criterion variable
- Condition 4. The effect of the
predictor variable must be less in equation three than in equation two.
Full
mediation is achieved when the predictor variable influences the criterion
through the mediator. In terms of the regression equation, the beta weight for
the predictor is significant in equation two but non-significant in equation
three when the mediator is controlled for. Partial mediation is achieved when
the predictor variable influences the criterion variable through the mediator
indirectly and directly. Baron and Kenny (1986) argue that partial mediation warrants
a conclusion of a mediation effect as it is unrealistic to eliminate the
relationship between the predictor variable and the criterion variable totally
(p.1176).
