Comparative Studies, Surveys, and Designed Experiments

Abstract

Investigations that can benefit from a statistical contribution come from a variety of sources. Some examples are provided below:

• Surveys involve the sampling of collections of people or objects to gather information on numerical properties of the collections. The surveys may range from opinion polls to judge voter behaviour or social preference in human populations, to measurements of pollution levels or environmental change. Surveys are observational studies.

• Experiments arise when investigators select subjects or experimental units and in some manner manipulate the sample members by applying treatments. Note that the term “treatment” is not limited to a medical application, such as giving a drug to a subject. It applies to any action that might affect the response of the subject or experimental unit, as, for example, a change in the environment in which the sample members are placed. In this chapter and following chapters we concentrate on experiments that involve the comparison of treatment effects, because this raises important issues of design of experiments so that comparisons can be made in an efficient manner and with the aim of avoiding confusing treatment effects and nontreatment effects. We reserve the term designed experiment for those experiments in which the sample members are randomly allocated among the treatments. The reason for distinguishing this subset of experiments is made apparent below.