Stratified Sampling

Abstract Stratified sampling is a probability sampling method that is implemented in sample surveys. The target population's elements are divided into distinct groups or strata where within each stratum the elements are similar to each other with respect to select characteristics of importance to the survey. Stratification is also used to increase the efficiency of a sample design with respect to survey costs and estimator precision. In this article, the foundations of stratified sampling are discussed in the framework of simple random sampling. Topics include the forming of the strata and optimal sample allocation among the strata. Practical implementation issues for stratified sampling are discussed and include systematic sampling, implicit stratification, and the construction of strata using modern software. The importance of using stratified sampling in practice is demonstrated by its usage in five major large‐scale health surveys conducted in the United States and the United Kingdom. For these surveys, details of the stratification and sampling methods are provided. Topics include multistage cluster sampling within strata and the use of systematic and probability proportional to size sampling.