This book is about the statistical aspects of sample surveys. However, it is mainly concerned with the statistics of samples that are not of the simplest kind and therefore an appreciation of the process of sample design and of the realities of sampling is an essential requirement for assessing the statistical approach necessary for any given survey. Before proceeding to the methods of dealing with data from samples which are not of the simplest and most straightforward kind, we must first understand why this additional complexity may be necessary, desirable or inevitable in the first place.

Accordingly, we devote this chapter to a brief review of the main methods of sampling and in particular to the thinking that underlies them. We do not intend to be comprehensive in scope or to be prescriptive. We shall cover only those issues affecting the statistical side of survey sampling, and not matters of data collection procedures, fieldwork administration or processing. For a fuller understanding there is a wide range of readily available texts on both the theory and the practice of sampling (see Appendix B for suggestions for further reading).

Survey sampling in the real world is frequently an imperfect process, and we must come to terms with its imperfections, not pretend that they do not exist.