I tried to post this twice already, and my browser keeps crashing
third times a charm!
A mediator is a variable through which one IV affects a DV. To be considered a true mediational model, you need longitudinal data. For example, lets say a safe sex program is predicted to increase rates of condom use. The developers of the program believe that it will increase condom use by increasing self-efficacy among teens. In this model, program (received or not) is the IV, self-efficacy is the mediator, and rate of condom use is the DV.
A moderator is a variable that affects the degree to which two other variables are related. For example, lets say that the association between rumination and depression is stronger for women than for men. In this model, sex (male,female) is a moderator of the rumination/depression association.
Although the model you posted above might adhere to the rules of mediation, it is cross-sectional and doesn't make much conceptual sense to interpret it as mediation. Rather, I think it is better conceptualized as a covariate analysis. In other words, the association between Native American status and medication adherence is better explained by the third variable of income. It doesnt make as much sense to think of Native American status influencing medication adherence through income.
Hope that helps!
