How to Choose a Sample Size

RESEARCH & DISSERTATION All Homework Solutions · expert essay, research and dissertation help

“How many participants do I need?” is one of the most common questions in student research. The honest answer depends on your design, but there is always a defensible way to decide.

Sample size affects whether your study can answer its question, how convincing your results are and whether your ethics application and proposal are approved. Quantitative and qualitative research approach the question very differently, and student projects also face practical limits of time and access. This guide explains the logic behind each approach and how to justify your sample size in a methodology chapter.

Start from the research question and design

Sample size follows from what you are trying to find out. A study estimating how common something is in a population needs enough participants for a precise estimate. A study testing whether two groups differ needs enough to detect a meaningful difference. A qualitative study exploring experience needs enough to capture the range and depth of that experience. Clarify your aim before calculating anything.

Stuck on your assignment? Our expert writers can help.

Get help →

Understand statistical power

In hypothesis-testing studies, power is the probability of detecting an effect if one truly exists. Conventionally, studies aim for 80 per cent power at a significance level of 0.05. Underpowered studies risk missing real effects and producing inconclusive results. Power depends on the expected effect size, the significance level, the variability in the data and the sample size.

Run a power analysis

A power analysis calculates the sample needed to detect an effect of a given size. Free software such as G*Power makes this straightforward. You will need to specify the statistical test, the expected effect size, the significance level and the desired power.

Effect size (Cohen’s d) Approximate total sample, two-group t-test, 80% power
Small (0.2) About 790
Medium (0.5) About 130
Large (0.8) About 52

The table shows why expected effect size matters so much: halving it roughly quadruples the sample needed.

Choose an expected effect size sensibly

The expected effect size is the hardest input to justify. Base it on previous studies in your area, a pilot study or the smallest effect that would be practically meaningful. Avoid choosing a large effect simply because it produces a feasible sample size; examiners notice. If evidence is limited, explain the uncertainty and your reasoning.

Calculate sample size for surveys and estimates

For descriptive surveys estimating a proportion or average, sample size depends on the margin of error you will accept and the confidence level. For a large population, around 385 responses gives a margin of error of about plus or minus five percentage points at 95 per cent confidence. Smaller populations need fewer responses, and online calculators handle the adjustment. Remember to inflate your target to allow for non-response.

Account for subgroups and complex analyses

If you plan to compare subgroups or run regression with several predictors, your sample must be large enough for each analysis, not just the total. Regression guidelines often suggest a minimum number of cases per predictor variable, and subgroup comparisons can need many more participants than the overall analysis. Plan for the most demanding analysis you intend to run.

Approach qualitative sample size differently

Qualitative research does not aim for statistical generalisation, so power calculations do not apply. Sample size depends on the depth of data from each participant, the diversity of the population, the scope of the question and the analytic approach. Many student interview studies involve somewhere between six and twenty participants, but there is no universal number.

Information power

One useful concept is information power: the more relevant information each participant holds for your question, the fewer participants you need.

Use saturation carefully

Saturation — the point at which new data adds little new insight — is often cited as a justification for qualitative sample size. It is useful, but it must be described honestly. Explain how you judged it, for example by tracking new codes across successive interviews. Some qualitative approaches, such as reflexive thematic analysis, question whether saturation is a meaningful criterion at all, so check its fit with your method.

Be realistic about practical constraints

Student projects have limited time, budget and access. If a power analysis calls for 300 participants and you can realistically recruit 80, you need to adjust: change the design, focus on a larger expected effect, or reframe the study as exploratory or a pilot. Be transparent about these trade-offs rather than presenting an underpowered study as definitive.

Plan for attrition and non-response

Not everyone who agrees to take part will complete the study. Estimate likely drop-out or non-response from similar studies and recruit more participants to compensate. For a longitudinal study expecting 20 per cent attrition, a required final sample of 100 means recruiting about 125.

Justify your sample size in writing

In your methodology, state your target sample size, explain how it was determined, cite the method or tool used, and acknowledge limitations. In your results, report the actual sample achieved and discuss any shortfall.

“A power analysis using G*Power indicated that 128 participants would provide 80% power to detect a medium effect (d = 0.5) at α = .05. Allowing for 15% non-response, 150 participants were invited.”

Common mistakes to avoid

  • Choosing a sample size by convenience with no justification.
  • Inflating the expected effect size to make the sample feasible.
  • Ignoring subgroup analyses when planning.
  • Claiming saturation without describing how it was judged.
  • Presenting an underpowered study’s null result as evidence of no effect.