Why Sample Size Is Not Just a Number

Researchers often ask for "the sample size" as if it is a single number waiting somewhere inside a formula.

But sample size is not just arithmetic. It is a methods decision.

The number depends on what you are trying to estimate, compare, test, or understand. It also depends on assumptions, design, resources, ethics, and analysis. A sample size that looks precise on paper can still be weak if the reasoning behind it is unclear.

This is why a good sample size section should not only present a number. It should explain the logic behind the number.

1. Start with the main objective

Sample size should be linked to the primary objective of the study.

Are you estimating a proportion? Comparing two groups? Measuring change over time? Testing an association? Estimating a mean? Assessing prevalence? Evaluating an intervention?

Each of these may require a different calculation.

For example, a study estimating the prevalence of contraceptive use requires different assumptions from a study comparing contraceptive use between two groups. A trial requires different thinking from a descriptive survey. A qualitative study requires a different form of justification altogether.

If the objective is unclear, the sample size calculation will also be unclear.

Before calculating anything, identify the main question the study must answer.

2. Understand the assumptions

Every sample size calculation relies on assumptions.

These may include:

• Confidence level
• Margin of error
• Expected prevalence
• Expected mean or standard deviation
• Effect size
• Statistical power
• Significance level
• Non-response rate
• Design effect
• Number of groups
• Clustering

These assumptions should not be guessed casually. They should come from prior studies, pilot data, programme data, reasonable estimates, or a clearly stated rationale.

For example, if you assume a prevalence of 50%, explain why. If you add a 10% non-response rate, explain why that is reasonable for your setting. If you use a design effect because of cluster sampling, say so.

A sample size calculation without explained assumptions is like a budget without unit costs. It may produce a number, but it does not build confidence.

3. Match the calculation to the design

Different study designs require different sample size logic.

A simple cross-sectional survey is not the same as a cluster survey. A cohort study is not the same as a case-control study. A randomized trial is not the same as a routine programme evaluation. A qualitative interview study is not justified using the same logic as a prevalence survey.

If the design includes clustering, such as schools, facilities, communities, or groups, the sample size may need to account for similarity within clusters. If the study has multiple groups, the sample size should reflect the comparisons being made. If the study involves follow-up, loss to follow-up should be considered.

The calculation should fit the design you are actually using, not the easiest formula to find online.

4. Think about feasibility

A statistically calculated sample size still needs to be feasible.

• Can you recruit the required number of participants?
• Can you supervise the data collection properly?
• Can you afford the fieldwork?
• Can the team complete the study within the timeline?
• Can the data be cleaned and analysed well?

A very large sample may look impressive, but if the team cannot implement it properly, data quality may suffer. A very small sample may be easier to manage, but it may not answer the study question convincingly.

Feasibility does not replace statistical reasoning. It complements it.

A good proposal explains both: the calculated sample size and the practical plan for reaching it.

5. Do not ignore non-response

Many studies calculate the minimum required sample and then forget that not everyone will participate, complete the questionnaire, or provide usable data.

Non-response should be anticipated. The percentage added will depend on the population, setting, recruitment process, study topic, and data collection mode.

For example, online surveys may require a different non-response assumption from facility-based recruitment. Follow-up studies may need to account for attrition. Sensitive topics may also affect participation.

Adding a non-response adjustment is not a decoration. It is part of realistic planning.

6. Qualitative sample size is different

For qualitative studies, sample size is usually justified differently. The focus is not statistical precision, but depth, relevance, variation, and the richness of information.

A qualitative sample size justification may refer to the study aim, participant diversity, expected depth of interviews, specificity of the research question, and the analytic approach. Some studies use concepts such as saturation or information power, but these should be applied thoughtfully rather than dropped into the methods section like seasoning.

The key is to explain why the proposed number and type of participants are sufficient to answer the qualitative question.

7. The sample size section should tell a story

A strong sample size section should answer four questions:

• What is the sample size based on?
• What assumptions were used?
• Why are those assumptions reasonable?
• How will the study reach the required number?

For example, instead of simply writing, "The sample size will be 384," explain the objective, formula or approach, assumptions, adjustment for non-response or design effect, and final target.

That explanation helps reviewers trust the number.

Final thought

Sample size is not just a number. It is a methods argument.

It tells reviewers whether your study question is clear, whether your design is appropriate, whether your assumptions are reasonable, and whether your plan is feasible.

The calculation matters. But the reasoning matters just as much.