Correlation association and causation

Whether we use an experimental study or an observational one, we must be careful when we say that the results show us one thing causes another. There may be a relationship without causality. In a correlation, example, as a goes up b also goes up, and when a falls b also falls. (In a negative correlation, example, as a goes up b goes down.) A correlation may be interesting but it does not prove that a causes b. The changes may be due to simple chance, some other factor may be the cause, the two factors may be reinforcing each other, or it may be that in fact b is causing a. In an association, the presence of one factor may often be linked to the presence of another factor, but again, this does not demonstrate causation. It may just be coincidence or they may both be caused by a further factor.

When we find a relationship between one factor and another we can try to evaluate our belief in causality against the framework given by Hill (1965):

IDevice IconActivity 5: Causality framework

Read the article by Hill [HILL, Austin Bradford. The environment and disease: association or causation? Bull World Health Organ. [online]. 2005, vol. 83, no. 10 [cited 20-06-2007], pp. 796-798. ISSN 0042-9686.] and then write in your logbook your own definitions of the terms in his framework for evaluating the existence of causality. What is Hill's philosophy regarding the application of these rules to assess whether a relationship is causal? Now open you log book and complete the activity.


IDevice Icon Activity 6: evaluating for causality

This activity asks you to use Hill's criteria to evaluate the likelihood of causality with respect to certain statements, while making notes on your thinking in your log book. Now open your log book and complete Activity 6..

iDevice iconControl Groups and Confounding effects

In an experimental study, one accepted way of trying to assess causality is to use a control group; a group of experimental subjects who are not exposed to the factor of interest. The experimenter assigns subjects to the experimental group or the control group and carries out the experimental process, often known as the treatment on the experimental group only. If conditions for the two groups are otherwise identical and a significant effect is observed in the experimental group, but not in the control group, it is likely that the treatment has caused the effect. Statistical techniques can be used to determine what constitutes a significant effect.

Without the control group, observing an effect would not tell us about causality, but with the control group, providing we can be certain the conditions were otherwise identical between the groups, it is easier to draw a conclusion of causality. The difficult part is in ensuring that we have considered all the factors that may cause the conditions between the groups to vary. This is the skill of good experimental design. If a condition does vary between the groups in an unknown or uncontrolled way, this is known as a confounding factor and can lead to a spurious conclusion of causality. Even in an experiment where we believe we have controlled every factor, any conclusion of causality should be evaluated against Hill's criteria, especially those of plausibility and coherence.

You have now been introduced to the main areas that you should think through in your research design regardless of your methodology. In the next section you will learn about the two different approaches; qualitative and quantitative and how to decide which kind of approach is going to suit your own research.