Causal Models

If one event has a recognisable probability of influencing another event, a causal relationship is commonly assumed between the two events, and the former is said to cause the latter. Causation is a concept about which there is great debate. However, no longer is a deterministic approach to causation considered essential for a causal effect to be ascribed, in so far as the second event is always preceded by the first event, and that the first event is always followed by the second event. It is sufficient for the relationship between the two events to be expressed in terms of probabilities. In order to attribute a causal relationship between event A and event B, three conditions are necessary, but not sufficient: (a) B must not precede A in time, (b) A and B must covary together to a recognisable degree, and (c) no alternative explanation accounts as well as or better for the covariation between A and B. Where more than two events are under consideration, the relationships between the events may be expressed in the form of a causal model, which may be submitted to tests of coherence and consistency, and the parameters of the model estimated to examine whether or not they are significantly different from zero.