Skip Ribbon Commands
Skip to main content

Causal Inference

Last modified at 9/15/2015 4:40 PM by Arnold Bosman

Association and Causation

Epidemiologists aim to draw conclusions on whether an observed association is one of cause and effect. Establishing this relation (causation/causality) is a difficult task. In fact the concept of cause itself continues to be debated as a philosophical matter in the scientific literature. In this section we explore what is meant by causation and encourage an open mind about causal inference.  

As "cause" can be used to give undue weight to an association, it is important  to consider and remember its meaning. One definition of cause is a "preceding event", condition or characteristic that leads to a given outcome at that time. The mechanism behind a cause can be divided into necessary and sufficient components

Bradford Hill set out nine viewpoint on causality:  strength of association,  consistency, specificity, temporality,  biological gradient, plausibility,  coherence, experimental evidence, and analogy. While these viewpoints are helpful when considering cause and effect,he insisted that “none of [his] nine viewpoints can bring indisputable evidence for or against the cause-and effect hypothesis”.  What they can do, with greater or lesser strength, is to help epidemiologists make up their minds on the fundamental question - Is there any other way of explaining the set of facts before them? Is there any other answer equally, or more, likely than cause and effect? 

It is important to keep in mind that most judgments of cause in epidemiology are tentative and should remain open to change with new evidence. It is important to be remain critical, to aim always for stronger evidence, and to keep an open mind. Checklists of causal criteria should not replace critical thinking.

 "The world is richer in associations than meanings, and it is the part of wisdom to differentiate the two." John Barth, novelist.

The concept of causal inference is related to, yet differs from statistical inference, which is described elsewhere.


EPIET Lectures:

Causal Inference