mediating variable

n. A mediating variable explains how or why two variables are related by specifying that the mediating variable transmits the effect of one variable to another variable. There is a temporal relation such that the initial variable is related to the mediating variable and then the mediating variable is related to the outcome variable. Mediating variables may be observed measures used in a statistical analysis, or they may be theoretical constructs that guide a research project. One of the difficulties of research on mediating variables is the development of accurate measures of theoretical mediating constructs.

Mediating variables are often considered once a relation between two variables is established. Here the mediating variable provides an interpretation of the relation between two variables by elucidating a process by which the initial variable is related to the mediating variable and the mediating variable is related to the outcome variable. Stimulus-organism- response models provided the first examples of mediating variables in psychology. These models hypothesized that a mediating variable in the organism translates the stimulus to the response. For example, mental processes mediate how exposure to information affects behavior. More recent applications of mediating variables are in treatment and prevention studies, where a manipulation is designed to change a mediating variable hypothesized to be causally related to the outcome variable. If the mediating variable is changed by the treatment program and the mediating variable is truly related to the outcome variable, then the program will affect the outcome variable. For example, a prevention program that changes norms about smoking may reduce smoking.

Mediating variable relations may be complicated. More than one mediating variable may explain the relation between the initial and outcome variable. A mediating process may consist of a long chain of variables that lead from the initial to the outcome variable. Mediating variables may be categorical (e.g., divorced or not) or continuous (e.g., attitudes, norms) variables. Mediation relations may differ at different time lags; for example, a mediation relation may be observed for daily measures but not for weekly measures.

The statistical assessment of mediating variables consists of statistical tests of the relation of the initial to mediating variable and the mediating variable to the outcome variable. Random assignment of participants to levels of the mediating variable helps clarify whether the mediating variable is related to the outcome variable. Generally, a program of research is necessary to provide convincing evidence for a mediating variable, including experiments, statistical analysis, clinical observations, and qualitative methods.

- DVM