Logic models can help evaluators and program staff understand the mechanisms through which their program supports behavior change. They illustrate, in graphical form, the hypothesized relationships between program components and behavior change, including any potential mediating variables. Logic models can assist with the design of evaluation instruments and methods by suggesting areas of inquiry and timing of assessment.
Mediation is the process by which the initial relationship between a given independent variable and a given dependent variable acts through a second independent variable. This is seen more easily with a diagram, which illustrates the relationships between variables in the theoretical model provided. Mediation analysis is the process through which researchers ascertain the mechanisms working in a given relationship.
In the context of theory-based interventions, mediation analysis can be particularly helpful in determining the mechanisms through which behavior change operates. It is one way to test the extent to which the program theory is supported by the evaluation data.