Problems of Statistical Inference in Health with Special Reference to the Cigarette Smoking and Lung Cancer Controversy

Abstract

Disagreement about the role of cigarette smoking in lung cancer is traced to three sources of confusion. In the first, there is lack of agreement as to the definition of terms and what is implied when a specific factor is said to be a cause of a particular disease. A suggested definition is offered to clarify one of the issues causing dissent. The second area of disagreement concerns the method of measuring the strength of the relationship in a 2 × 2 table, and whether the relationship is more pronounced in one disease than another. There is no solution to this problem until the specific question to be answered is stated in detail. Slightly different questions can lead to radically different answers. The third area involves the making of decisions and the subsequent formulation of actions based upon existing data. In the present context, four specific action-type decisions are involved and the corresponding relative weight of evidence which is needed to substantiate each of them is considered.