Population

All definitions of the population start with a definition of an individual, the elementary units about which inferences are to be drawn. The population is then the collection or aggregation of the individuals or other elements about which inferences are to be made.

In statistical usage, a population is any finite or infinite collection of individual elements. That inferences are to be made to the collection of individuals is implied by this definition. The term universe is also used in statistics for infinite or hypothetically infinite set of elements.

In survey usage, a population is strictly a finite collection of the units from which information is sought in the survey, with additional specification. The term universe is avoided, because it implies the infinite and hypothetical; survey research is materially oriented.

There are dimensions of the definition required for implementation purposes. Survey research also often defines populations that are mixtures of units, often where the different types of units are hierarchically ordered. For example, a survey could be designed to collect data from schoolchildren, teachers, and schools in the same survey. All three types of units are part of the population, an aggregate of different types of units.

A population description in survey usage includes the content, the units, the extent, and a temporal dimension. For example, a survey of student performance in a national measurement of educational attainment could specify the population as all students (content), grouped in classrooms (units), in schools in the United States and its territories (extent), in 2008 (temporal dimension). This population is clearly countable and potentially subject to census rather than survey collection.

A population definition often must be modified to meet operational constraints, leading to several types of populations that must be specified. There are different overlapping terms applied to these populations. A population of inference may be constrained by operational considerations to a target population. For example, the resources available to the survey may require the elimination of elements that are costly or impractical to reach. A target population may consist of all students (content) in public schools (units) in the 50 states, the District of Columbia, and Puerto Rico (extent), enrolled at the time of data collection during the 2007–2008 school year (temporal).

Further restrictions arise as materials are obtained to select the sample. The collection of materials used to select the sample is referred to as a frame, and the aggregate of units in the frame is sometimes referred to as the frame population. For example, the frame may include only schools known to be in operation at the beginning of the school year; schools opened during that year are excluded.

Some survey researchers define yet another reduction of the population that accounts for nonresponse. For example, not all students may respond to a request to provide information for the survey; students may be [Page 591]absent repeatedly on survey administration days at a school; or parental permission was not obtained for all students selected for study. A survey population is sometimes used to describe this latter collection of individuals, implying all elements that would have responded to the survey request rather than only those in the sample that did respond. In principle, the population of inference, the target population, and the frame population can be delineated and counted outside the framework of the survey. The survey population is delineated within the conduct of the survey, less precisely.

...