ABSTRACT
This chapter describes the term "stratification", which refers to mutually exclusive and exhaustive subgroups within the population, but useful and efficient stratification will depend upon more than just the homogeneity of the strata with respect to the outcome variable. It presents the goals for creating strata and several methods for creating strata. Schouten and Shlomo discuss how to approach the identification of strata using Partial R-Indicators with a "structured trial-and-error approach". The National Survey of Family Growth (NSFG) experimented with using regression diagnostics to define strata. The auxiliary data used in this study include data from the sampling frame, paradata, and data from the screening interview. A static ASD is considered in which the population is stratified beforehand based on linked administrative data. Refining and testing strata over time may lead to improvements in outcomes. Given that this step is critical for the success of ASD, resources used in the evaluation of various stratification schemes will not be wasted.
