Once you decide the general sampling method--random, systematic, stratified, convenience-- you can determine the sampling frame or frames needed to identify all people within the sample population.
A sampling frame can be thought of as any source of data that may be used for identifying households or individuals. For example, all of the following sources of individuals or households are sampling frames:
- Listed telephone numbers with a given area code
- A list of all students in the 8th grade at a given middle school
- Addresses of all households listed in a census within a set of geographic coordinates.
A sampling frame is almost never a perfect match of the target population. Imperfection occurs when there are elements of the target population that are not included in a given sampling frame. The fit of a sampling frame to the target population is referred to as coverage.
In a telephone survey, for example, a listed landline telephone frame would not include nearly 40% of the population who are either cell-phone only or cell-phone mostly users, and this percentage would be even higher for a young adult target population. A cell-phone frame would need to be used in addition to a landline frame in order to obtain better survey coverage and a better representation of the target population in the sample population. Many surveys use multiple sampling frames to ensure greater sampling coverage.
Sampling frames need to be carefully considered to avoid potential coverage errors, including undercoverage, duplication, or ineligible units.
Maximizing survey coverage also depends on choosing the appropriate survey mode, and in practice, choice of survey mode and selection of sampling frame are interdependent.
The survey mode describes the method of contact with individuals and how information (data) is collected. The most common survey modes are face-to-face, web, telephone, or mail. Selection of mode depends on many factors including:
- Degree of interviewer involvement needed
- Desired amount of interaction with the respondent
- Sensitivity of the subject matter or degree of respondent privacy needed
- How best to communicate survey questions to respondents
- Use of technology
Data collection methods in general, describing the instruments used to gather information from or about respondents, are not discussed here but are covered in Data Collection.
Citations and Further Reading:
Selvin, S. (1996). Statistical Analysis of Epidemiologic Data. New York: Oxford University Press.