ABSTRACT
This chapter deals with a description of the National Crime Victimization Survey (NCVS) design, and with parallels between survey samples and designed experiments. The NCVS is a large national survey administered by the United States (US) Bureau of Justice Statistics with interviews conducted by the US Census Bureau. Like the Population Survey, the NCVS follows a stratified, multistage cluster design. The NCVS design is similar to that of many other large government surveys: most have similar methods of stratification, clustering, and ratio estimation. Sampling weights and design effects are commonly used in complex surveys to simplify matters. Numerous parallels between sample surveys and designed experiments are discussed in H. Fienberg and J. M. Tanur and F. Yates. Randomization serves similar purposes in sampling and in designing experiments. Both design of experiments and sampling are involved in similar debates between using a randomization theory approach or using a model-based approach.
