Mediation Analysis

Statistical Analysis

A mediational hypothesis is a kind of statistical test that assumes that the relationship between an independent and dependent variable is mediated by (or accounted for by) a third variable. This is what is known as a mediating variable. The goal of the mediating variable is to completely mediate the relationship so that when the variable is added to a model, the independent variable does not affect the dependent variable.

Researchers Baron and Kenny list out a series of four required steps to follow when running a mediational hypothesis. Step one is to determine that the initial variable is correlated with the outcome variable. This is important because it establishes that there is an effect for a third variable to mediate. Similarly, steps two and three are related to determine that there is a correlation between the initial variable and the mediator variable and between the outcome variable and the mediator variable. Finally, the last step is to establish the complete mediation across the variables. Meaning that the effect of the initial variable on the outcome variable will be zero when the mediation variable is added.

If the data satisfies all four of these steps, then according to the researchers, the data is reliable for a mediational hypothesis to work. It is important to note, however, that the completion of these steps does not mean that mediation has occurred, only that a mediation is one plausible model that would fit the data.

request a consultation
Get Your Dissertation Approved

We work with graduate students every day and know what it takes to get your research approved.

  • Address committee feedback
  • Roadmap to completion
  • Understand your needs and timeframe