Intermediate Causal Factors

Heuristically, an intermediate causal factor, or causal mediator, is a variable that is causally influenced by the exposure and in turn causally affects the outcome of interest. Thus, the mediator is in some sense partially responsible for the total causal effect of exposure. The problem is to articulate exactly what this idea of partial causation really means. As discussed previously, classical statistical theory has no clear definition of causation in general, let alone in this more complex situation. Consequently, there has been much confusion regarding whether and how a total causal effect can be parsed into components that reflect the direct and indirect effects of exposure.


Many researchers believe that it is possible to resolve the total causal effect into direct and indirect components, using variations of ordinary statistical models (Baron and Kenny, 1986; MacKinnon et al., 2002). In essence, this is attempted by adjusting for the intermediate variable to estimate a direct (or net) effect, and then subtracting this direct effect from the total effect, to derive the indirect effect. For example, it is known that some nonsteroidal anti-inflammatory drugs (NSAIDs) can increase blood pressure. Suppose we are studying the effect of an NSAID on myocardial infarction (MI). An increase in blood pressure raises the risk of an MI. Ideally, we would like to know the direct effect of the NSAID, as if it had no effect ...

Get Bias and Causation: Models and Judgment for Valid Comparisons now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.