Causal Inference

In an attempt to define causationCausation, we came to realize that there are levels to consider. Necessary causes are those that must be present to see the effect. Sufficient causes, those that can produce the effect are almost always a collection of component causesComponent causes acting together. Most frequently, there are different collections of component causes that are sufficient to produce the effect. The frequency with which a particular component cause occurs in sufficient causes influences the apparent strength of association between that component cause and the effect. When interpreting an observed association, we realize there are three possible explanations. One is that it reflects a causal relationship, but there are two other possible explanations. These are that the association reflects a biased relationship or that the association reflects a chance relationship. To decide if an association is due to a causal relationship, we eliminate bias and chance as likely explanations. Cause is concluded through the process of elimination. To assist in this evaluation, Sir Austin Bradford Hill suggested nine criteria for causationCausation. For the most part, these help to evaluate bias and chance as likely explanations for an observed association.