Home Statistical Dictionary Sampling Frame
Sampling Frame
The actual list, register, or mechanism used to identify and select units from the target population -- the operational source from which a sample is physically drawn.
In Plain English
The sampling frame is the concrete, physical thing you actually sample from, a phone directory, a voter registration list, a customer database, a map of city blocks, as opposed to the target population, which is the abstract group you'd ideally like to study. In a perfect world the frame and the target population would be identical, but in practice a frame almost always misses some members of the population (undercoverage) or includes units that don't belong (overcoverage), and understanding exactly how the frame diverges from the population it's meant to represent is one of the first, most consequential steps in evaluating whether a survey's results can be trusted.
Definition
The sampling frame is the list, register, database, or other enumerable mechanism from which sample units are actually selected, serving as the operational stand-in for the target population. Discrepancies between the frame and the target population produce coverage error, undercoverage occurs when population members are missing from the frame entirely (and so have zero probability of selection, no matter how the sample is drawn), overcoverage occurs when the frame includes ineligible units or duplicates, and multiplicity occurs when a single population unit appears more than once in the frame with unequal or unaccounted-for selection probability; a well-constructed sampling frame is a prerequisite for any probability sampling design, since design-based inference relies on knowing every unit's inclusion probability, which is impossible for units the frame never gave a chance of selection.
Properties
- A sampling frame's coverage of the target population is a design-time property that cannot be fixed by statistical adjustment after the fact, units entirely missing from the frame had zero probability of selection and are structurally excluded from the sample, no reweighting scheme can recover information about people who were never eligible to be drawn.
- Common real-world sampling frames -- telephone directories, address lists, voter registries, membership databases -- each have characteristic coverage gaps (unlisted numbers, recent movers, non-voters, non-members), which is why survey methodologists routinely report and discuss frame coverage as a core piece of a study's documented limitations.
- Multiplicity, where a single population unit corresponds to more than one entry in the frame (a person with two phone numbers, a household appearing on two overlapping lists), inflates that unit's true selection probability above what the sampling design assumes, and must be corrected for, typically by adjusting weights inversely to the number of frame entries linked to each unit.
At a Glance