Population

In research, a population is the entire group of individuals, objects, events, or measurements about which a researcher wishes to draw conclusions. It is the complete set of entities that share a defined characteristic and to which the study's findings are intended to apply.

Nearly every quantitative study begins with a question about a population, yet researchers almost never have the resources to study every member of the population. Instead, they study a sample and use statistical methods to generalize findings to the population from which the sample was drawn. The validity of this generalization depends on how well the population is defined, how the sample is drawn from it, and how well the achieved sample represents it. Vague or unstated population definitions make it impossible to determine to whom the findings apply, while poorly drawn samples produce biased estimates of population parameters. Clearly defining the target population (the population to which the researcher wants to generalize) and the accessible population (the population from which the sample is actually drawn) is a necessary first step in any study that aims to generalize its findings, shaping decisions about sampling strategy, inclusion and exclusion criteria, recruitment methods, and the scope of the conclusions.

Populations are defined by a set of inclusion and exclusion criteria that specify who or what belongs to the population. In clinical research, these criteria typically include demographic characteristics (age, sex), clinical characteristics (diagnosis, disease severity, comorbidities), geographic characteristics (country, region, care setting), and temporal characteristics (time period during which participants are eligible).

The target population is the broadest group to which the researcher wants the findings to apply. For a study of a new diabetes medication, the target population might be all adults with type 2 diabetes worldwide. The accessible population is the subset of the target population that the researcher can realistically access, such as adults with type 2 diabetes who attend clinics in a particular health system during the recruitment period. The study sample is the subset of the accessible population that actually participates in the study.

The gap between the target population and the study sample introduces threats to external validity (generalizability). If the accessible population differs from the target population in important ways (for example, if only patients at academic medical centers are accessible, but the target population includes all care settings), the findings may not generalize. If the sample is a biased subset of the accessible population (for example, if only volunteers who respond to advertisements participate), additional generalizability concerns arise.

In statistical theory, the population is the source from which samples are drawn, and population parameters (such as the population mean, population variance, and population proportion) are the quantities that researchers seek to estimate using sample statistics. The sampling distribution, a central concept in inferential statistics, describes how a sample statistic would vary across all possible samples of a given size drawn from the population. This distribution enables the construction of confidence intervals and hypothesis tests that link sample findings to population parameters.

Finite populations have a countable number of members, N, (such as all students enrolled at a university in a given year), while infinite or conceptual populations are theoretically unlimited (such as all possible patients with a given condition, past and future). The distinction matters for some statistical calculations, such as the finite population correction, which multiplies the standard error by the square root of (N − n) ÷ (N − 1) when a sample of n is drawn without replacement: for a sample of 2,000 from 10,000 students it shrinks the standard error by about 11%, while for a sample that is a small fraction of the population it changes almost nothing. It does not change the fundamental logic of statistical inference.

Define your target population explicitly at the beginning of every study. State inclusion and exclusion criteria clearly in the methods section. Describe the accessible population and the sampling strategy used to draw the study sample. Discuss the representativeness of the sample relative to both the accessible and target populations in the limitations section. Use population parameters as the reference quantities for all statistical inference, and be explicit about the population to which your conclusions apply.

A common mistake is failing to define the population clearly, leaving readers unable to determine to whom the findings apply. "We studied patients with heart failure" does not specify the age range, severity, care setting, or geographic scope. Another error is generalizing findings beyond the population from which the sample was drawn without acknowledging the inferential leap. A study conducted entirely in hospitalized patients cannot automatically be generalized to outpatients, and a study of volunteers may not reflect the experiences of non volunteers. Confusing the study sample with the population is a further mistake: the sample is a tool for learning about the population, not the object of interest itself. Finally, using convenience samples without discussing the limitations of this approach for population inference is a common shortcoming in many fields.

A national health survey aims to estimate the prevalence of hypertension in U.S. adults. The target population is all noninstitutionalized civilian adults aged 18 and older residing in the United States. The accessible population is the part of this group that the survey's sampling frame covers and that can be located and contacted through its multistage probability sampling design, which selects counties, then census blocks, then households, then individuals within selected households. The study sample comprises the 5,500 adults who complete the health examination, including blood pressure measurement. Because the sampling design gives every adult in the households it covers a known, nonzero probability of selection, the sample estimates can be generalized to the target population with quantifiable sampling error. Survey weights adjust for differential probabilities of selection and for nonresponse, but they remove nonresponse and coverage bias only to the extent that the characteristics used to build them capture how the missing adults differ from those examined.