ABSTRACT
Medicine is an empirical science, and decisions are almost invariably based on the experience gained on a sample of subjects. Different samples contain different individuals. Since individuals differ from one another, it is natural that the samples also vary from one another. For example, it is possible in a small therapeutic trial of a new regimen that six patients respond in a rst sample of 10, three in a second sample, and seven in a third sample of the same size. Similarly, mean reduction in cholesterol level in a rst sample of 15 hypertensive-hypercholesterolemic subjects may be 30 mg/dL, in a second sample 24 mg/dL, and in a third sample 25 mg/dL after the same therapy. Different samples tend to give different results, and this generates considerable uncertainty. One of the main objectives of statistical methods is to study these sampling uctuations and develop strategies to draw valid conclusions for the target population on the basis of just one sample.
