Kerry and Bland (1998) describe the analysis of cluster randomized studies where a group of subjects are randomized to the same treatment. For ex-

ample, when women in some randomly selected districts are oered breast cancer screening while women in other districts are not oered the screen-

ing, or when some general practitioners are randomly assigned to receive one or more courses of special training and the others are not oered the training. The response (some characteristic of the patients) in the clus-

ter trials must be measured on each patient rather than the group as a whole. What is the experimental unit in this type of study? How would you describe the individual measurements on patients?