Many of the datasets archived in SAMHDA are designed to represent particular populations. For instance, the National Survey on Drug Use and Health (NSDUH) is designed to represent citizens of the United States, ages 12 and up. Since not every citizen can be surveyed, such studies utilize a stratified random sampling strategy. That means that the population is divided into broader groups (called sampling units) from which individual persons are randomly selected for participation. Depending on the focus of the research, more participants may be selected from particular sampling units than is necessary for proportional representation (a practice known as oversampling). This is done to ensure that the dataset will have an adequate number of cases to facilitate analysis of the subpopulation. Mathematical weights are used to adjust the findings to be proportionally accurate; thus, weights are adjustments used to ensure that findings accurately represent the population. Users should always consult the codebook to determine which weight, if any, is required.
No comments:
Post a Comment