How Should You Design a Study When the Entire Eligible Population Is Small?

What If Your Entire Research Population Is Smaller Than the Recommended Sample Size?

Suppose you want to investigate the experiences and academic outcomes of students enrolled in a highly specialized university program. Only 38 students meet the eligibility criteria, and there is no comparable program elsewhere in the institution.

You conduct a conventional sample size calculation and discover that the suggested number of participants exceeds the entire eligible population. Recruiting more students is impossible without changing the population or research question.

Should you abandon the study, recruit everyone, use qualitative methods instead, or proceed with quantitative analysis despite the small number of participants?

This problem also arises in research involving rare conditions, specialized professionals, small organizations, isolated communities, and narrowly defined educational settings.

The challenge is not simply obtaining a larger sample. It is designing a study whose scientific objectives, measurements, and analytical claims remain appropriate when the population itself is inherently limited.

Design Around the Available Population and the Question You Can Credibly Answer

How to Build a Defensible Research Design for a Small Eligible Population

What does it mean for the entire eligible population to be small?

A small eligible population exists when relatively few individuals, institutions, events, or other units satisfy the study's inclusion criteria within the defined population and time frame.

For example, a university may have only 25 students enrolled in a specialized doctoral program, while a national professional association may have only 70 members with a particular qualification.

Population size is not the same as sample size. The population comprises all eligible units relevant to the research question, while the sample comprises the units actually observed or selected.

If 38 students are eligible and 30 participate, the eligible population is 38 and the participating sample is 30.

Researchers should distinguish a genuinely small population from a larger population that is difficult to access. These situations may require different recruitment and inferential strategies.

Why might conventional sample size calculations be inappropriate?

Sample size calculations depend on the research objective, sampling design, target effect or precision, statistical model, and assumptions about the population.

A conventional formula may assume sampling from a much larger population or detecting a specified effect with predetermined power.

When the eligible population contains fewer individuals than the calculated requirement, the result does not establish that those individuals are unsuitable for research.

Instead, it indicates that the proposed statistical objective may not be achievable under the stated assumptions.

For example, a study requiring 100 participants to detect a small difference with 80% power cannot satisfy that requirement by recruiting all 38 eligible individuals.

Researchers should reconsider the target precision, estimand, design, or scientific objective rather than claiming that a census automatically solves the power problem.

Why must researchers distinguish finite-population and broader-population inference?

A finite-population study concerns a specifically defined collection of eligible units, such as all students enrolled in a particular program during one academic year.

A broader-population study seeks conclusions about individuals or settings beyond those observed, such as future students entering similar programs.

The distinction matters because observing every member of a finite population can eliminate sampling uncertainty about a fully observed descriptive quantity for that particular population.

However, it does not eliminate measurement error, nonresponse, temporal variability, or uncertainty about future or unobserved populations.

Nor does a complete census establish causal effects merely because every eligible person was included.

Researchers should identify which type of inference the research question requires before selecting a sample size formula or statistical test.

Should researchers attempt to recruit everyone?

Recruiting the entire eligible population may be sensible when the population is small, accessible, and participation can be requested ethically.

This approach is commonly described as a census or complete enumeration when all eligible units are included.

It can maximize coverage of the defined population and reduce uncertainty arising from selecting only a subset.

However, inviting everyone does not guarantee complete participation. Refusal, incomplete measurements, and withdrawal may still occur.

Moreover, participation may be associated with the outcomes under investigation, creating nonresponse bias even when all eligible individuals were invited.

The detailed decision is addressed in whether researchers should recruit the entire eligible population.

When is a descriptive design appropriate?

Descriptive research may be particularly valuable when the objective is to characterize a small, clearly defined population.

For example, researchers may document the educational backgrounds, learning experiences, or academic outcomes of all students enrolled in a specialized program.

Counts, proportions, distributions, and individual-level variation may be more informative than statistical significance testing.

If the entire finite population is observed, descriptive statistics can directly characterize that population during the specified period.

However, researchers should not assume that the same descriptive results necessarily apply to future cohorts or other institutions.

The value of the study depends on the relevance and accuracy of the description, not whether it produces a statistically significant result.

Can small-population research use quantitative methods?

Yes. Quantitative research is not defined by having a large sample.

Researchers can calculate descriptive statistics, estimate relationships, analyze repeated measurements, and use statistical models when the data and assumptions support those methods.

However, small samples may produce unstable estimates, wide confidence intervals, and difficulties fitting complex models.

For example, a regression model containing many predictors may be poorly supported by observations from only 25 participants.

Researchers should avoid using an excessive number of parameters relative to the information available.

The important question is whether the chosen method can provide interpretable estimates for the intended objective.

Further considerations are addressed in using quantitative methods with small populations.

When might qualitative research be preferable?

Qualitative research may be appropriate when the objective concerns experiences, meanings, processes, or contextual explanations rather than numerical population parameters.

For example, interviews with members of a small specialized professional community may provide detailed information about how institutional practices shape their work.

However, qualitative research should not be chosen merely because a quantitative sample size target cannot be met.

Qualitative approaches have their own methodological requirements concerning sampling, data richness, analytic coherence, reflexivity, and the adequacy of evidence.

A small population does not automatically make qualitative inquiry superior.

The research question should determine the methodological approach.

Can mixed methods be useful?

Mixed-methods designs may be appropriate when numerical patterns and contextual explanations are both necessary to answer the research question.

For example, researchers could describe academic outcomes for all students in a specialized program and conduct interviews to understand experiences that may help explain those patterns.

However, adding qualitative interviews does not statistically compensate for inadequate quantitative power.

Similarly, descriptive quantitative results do not automatically validate qualitative interpretations.

The components should be integrated through a clear methodological rationale, with explicit attention to what each contributes.

Mixed methods should be selected for complementary inferential purposes rather than as a general remedy for small samples.

Can repeated measurements increase the information available?

Repeated measurements may provide additional information about individual trajectories, within-person changes, and responses to interventions.

For example, researchers studying 20 students could measure academic engagement weekly rather than collecting only one observation per student.

However, repeated observations from the same participant are usually correlated.

Twenty participants measured ten times do not constitute 200 independent participants.

The analysis should account for within-person dependence and the timing of measurements.

Repeated measurement may improve precision for particular within-person estimands, but it cannot automatically replace the between-person information required for other questions.

The design implications are examined in using repeated measures in small populations.

Can crossover designs improve efficiency?

A crossover design exposes participants to more than one intervention condition in a specified sequence, often using randomization to determine treatment order.

Participants may serve as their own controls, potentially reducing between-person variability.

This can be useful when treatment effects are reversible, outcomes can be measured repeatedly, and carryover effects can be managed.

However, crossover designs may be inappropriate for permanent interventions, irreversible outcomes, or treatments with lasting effects.

For example, a short-term instructional interface may be suitable for a crossover comparison under carefully controlled conditions, while an intervention intended to produce lasting learning may create substantial carryover.

The design should be selected according to the intervention and outcome rather than solely because the population is small.

When are single-case experimental designs appropriate?

Single-case experimental designs use repeated measurements and planned manipulation of intervention conditions to investigate functional relationships within individuals or small sets of cases.

Examples include multiple-baseline and withdrawal designs.

They may be appropriate when the research objective concerns the effect of an intervention on repeatedly measurable behavior or performance.

However, a descriptive case study is not automatically a single-case experiment.

Experimental control, repeated observation, and appropriate replication are central to the methodology.

Researchers should evaluate whether the intervention can be introduced, delayed, alternated, or withdrawn ethically and meaningfully.

These designs are not substitutes for population-level effect estimation when that is the actual research objective.

Can Bayesian methods help with small populations?

Bayesian methods combine observed data with specified prior information to produce posterior distributions.

They may be useful when researchers have credible external evidence or need to express uncertainty directly through probabilistic estimates.

However, when few observations are available, results may be strongly influenced by prior assumptions.

Researchers should evaluate prior sensitivity and distinguish information contributed by the current data from information introduced through the prior.

Bayesian methods do not automatically make a small population statistically sufficient for every objective.

What role do exact and small-sample methods play?

Some conventional statistical approximations perform poorly when sample sizes or event counts are small.

Exact tests, permutation procedures, penalized estimators, and small-sample corrections may be appropriate in selected settings.

However, their assumptions and interpretation depend on the design.

For example, a randomization-based test requires an appropriate randomization structure, while an exact test may rely on particular distributional or conditioning assumptions.

These methods can improve inferential calibration under certain conditions, but they do not guarantee useful precision or correct selection bias.

The analytical method should be planned alongside the research design.

Why can a small population create confidentiality risks?

Participants in small or distinctive populations may be identifiable even when names and direct identifiers are removed.

For example, reporting the exact position, age, institution, and experience of one of only three eligible professionals may permit identification.

Researchers should evaluate deductive disclosure risks, particularly when publishing detailed tables or qualitative quotations.

Suppressing small cells, aggregating characteristics, or limiting potentially identifying contextual information may be appropriate.

However, privacy protections should not be applied mechanically if they distort the scientific interpretation.

Ethics approval, consent arrangements, and applicable data-protection requirements should guide collection and reporting.

How should researchers interpret generalizability?

Generalizability depends on the relationship between the observed population and the population to which conclusions are intended to apply.

Observing every member of one specialized program may support an accurate description of that program during the study period.

However, extending those findings to other programs requires evidence about relevant similarities and differences.

In qualitative research, transferability may be supported through detailed contextual description and transparent interpretation rather than statistical representativeness.

In experimental research, replication across individuals, settings, or occasions may strengthen claims about broader applicability.

Researchers should not equate small population size with an absence of scientific contribution.

The appropriate scope of inference should be stated clearly.

What if the population is too small for the original hypothesis?

Researchers should distinguish a scientifically valuable question from a particular statistical formulation of that question.

If the available population cannot support a precise estimate of a small effect, the study may still provide useful descriptive estimates, exploratory evidence, or detailed within-person information.

However, changing the objective after examining results can create selective reporting concerns.

Researchers should establish realistic objectives before data collection and document substantive protocol revisions transparently.

When the original question genuinely requires information that cannot be obtained, the appropriate response may be to narrow the estimand, collaborate across sites, extend recruitment, or reconsider the study.

A small population should influence the design from the beginning rather than become an explanation introduced only after an inconclusive analysis.

Designing a Study of a Specialized Program with Only 38 Eligible Students

The study may provide useful evidence about the observed students and potentially a strong description of the defined program population.

However, the four nonparticipants may differ from those observed, and the small number of participants may limit the precision of subgroup comparisons.

The researchers should not claim that recruiting nearly everyone guarantees adequate power or establishes a causal effect of participation.

Common Misconceptions About Research with Small Populations

Match the Research Objective to the Information the Population Can Provide

Begin by defining the eligible population and clarifying whether the research objective concerns description, association, prediction, or causal effects.

Then determine whether complete recruitment, repeated measurement, an experimental arrangement, or a different analytical objective can provide useful evidence.

When the eligible population is small because the condition or exposure itself is rare, review how to study rare populations when conventional sample size requirements are impossible to meet.

The strongest design is not necessarily the one that imitates a large-population study. It is the one whose claims remain proportionate to the evidence that can actually be obtained.

Before Finalizing a Small-Population Research Design

Frequently Asked Questions About Small-Population Research

A Small Population Requires an Appropriate Design, Not an Artificially Larger Sample

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.