ABSTRACT
Stratified samples first create homogeneous groups from the population, and then make a proportional but random selection from each of the groups. This can help with several difficulties found in the previous methods, especially improving the representation of specific groups from the population. To obtain a stratified random sample, first divide a population into strata. For example, researchers can easily divide a population into men and women because gender is typically identifiable. Researchers are not confined to using just two variables for stratification. They can further increase precision by using multiple strata to select a given sample. Having a larger number of strata is only better with variables that are both relevant and independent of each other. To see what is meant by independence, suppose a researcher stratifies on both age and number of years employed. Because older individuals have had more years in which to be employed, the two variables tend to be highly correlated.
