Research work · 2003

Inference 2: causation

Robert Stewart

Practical Psychiatric Epidemiology

Abstract

The study of cause and effect forms the basis for most human interaction. The repetitive investigation of actions and their consequences can be readily seen in children's behaviour. Adult behaviour may be more complex but essentially involves identical principles. When we speak to someone for the first time, an initial impression is formed. If the conversation proceeds, the impression (hypothesis) is tested and refined through evaluating actions (what we say) and their consequences (the reaction or reply this provokes). If an unknown factor is present (e.g. the other person is preoccupied with something else), the relationship between cause and effect may be misinterpreted resulting in a false impression (e.g. that they are rude or unfriendly). The process can be seen as a repeated series of experiments, albeit unconscious. All of us are therefore involved in active cause–effect research for most of our waking lives. However the inferences (whether true or false) derived from these day-to-day experiments apply only to ourselves. Science and philosophy on the other hand seek to uncover truths that are generalizable beyond the individual. Because of this, their experiments require greater scrutiny. Research may be divided into that which is observational (describing what is there) and that which is analytic (explaining why it is there). Deducing cause and effect relationships is central to analytic research. The ‘result’ of any given experiment is indisputable. What is open to interpretation is what caused that result. As discussed in Chapter 12, a series of questions have to be asked. What is the likelihood of it having occurred by chance? Was it caused by problems in the design of the study (bias), by the influence of a different factor to that hypothesised (confounding), or by a cause–effect relationship in the opposite direction to that anticipated (‘reverse’ causality)? If the anticipated cause–effect relationship is supported, what precise cause and effect were being measured in the study under consideration and how might other factors contribute to this? And what are the implications of the findings? The focus for critiquing a research report (apart from allegations of deliberate falsification) strictly speaking should not be the reported ‘Results’ but the ‘Discussion and Conclusions’—the inferences (particularly regarding cause and effect) which can be drawn from the results and therefore the generalisability of findings beyond the experimental situation.

DOI: 10.1093/med/9780198515517.003.0013