5. Designing a Sample
Many other forms of sampling exist to address special situations such as studying rare or less common populations. For example, adaptive and network sample designs involve selecting a probability sample of population units (e.g., persons) and collecting data about the presence of a relatively uncommon characteristic (e.g., whether the respondent has a rare disease, whether the respondent engages in drug use). If the uncommon characteristic is detected, then a list of additional units who might also have this trait is elicited from the respondent, and the same data are collected on all or a sample of those units.
This is another example of oversampling part of the population, and the procedures should be designed to provide information to calculate the selection probability and survey weight for each unit. Methods such as respondent-driven sampling often do not provide this kind of information and should be avoided.