How to Calculate Sample Size for a Cross-Sectional Study
Step-by-step sample size for a cross-sectional (prevalence) study: precision, finite population correction, design effect and non-response.
Cross-sectional studies are the most common design in student research, especially prevalence and knowledge–attitude–practice (KAP) studies. Their sample size calculation is straightforward once you understand four choices: the expected prevalence, the precision, any correction for a small population and any adjustment for cluster sampling.
Step 1: Find the expected prevalence (p)
Search for a study in a similar population and setting. If several studies exist, use one that is recent and closest to your setting, and cite it. If no estimate is available, using p = 0.5 gives the largest (most conservative) sample size.
Step 2: Choose the precision (d)
Precision is how close you want your estimate to be to the true value.
- Absolute precision: a fixed margin, such as ±5 percentage points.
- Relative precision: a margin expressed as a percentage of p, such as 20% of p.
Absolute precision of 5% suits prevalences between roughly 20% and 80%. For a low prevalence such as 4%, ±5% would be meaningless (the interval could include zero), so use a smaller absolute precision or relative precision.
Step 3: Calculate
n = Z² × p × (1 − p) / d² (Z = 1.96 for 95% confidence)
| Scenario | Calculation | n |
|---|---|---|
| p = 30%, d = 5% absolute | 3.8416 × 0.30 × 0.70 / 0.0025 | 323 |
| p = 30%, d = 20% of p (= 6%) | 3.8416 × 0.21 / 0.0036 | 225 |
| p unknown (use 50%), d = 5% | 3.8416 × 0.25 / 0.0025 | 385 |
Step 4: Adjust for a small, known population (optional)
If you are sampling from a small, defined population (for example, all 1,000 nursing students in one college), you can apply the finite population correction:
n(adjusted) = n / [1 + (n − 1) / N]
With n = 323 and N = 1,000: 323 / (1 + 322/1000) = 323 / 1.322 = 244.3, so 245.
Step 5: Adjust for cluster sampling (if used)
If you sample clusters (villages, schools, wards) rather than individuals, people within a cluster tend to be similar. Multiply the sample size by the design effect, often taken as 2 when no better estimate exists. With n = 323 and a design effect of 2, you need 646.
Step 6: Add a non-response allowance
Final n = n / (1 − expected non-response rate)
With 10% non-response: 323 / 0.9 = 358.9, so 359.
Worked example in synopsis language
Taking the prevalence of hypertension among adults as 30% (reference), with an absolute precision of 5% and 95% confidence, the minimum sample size was 323 using n = Z²p(1 − p)/d². Allowing for 10% non-response, the final sample size is 359.
Common mistakes
- Using 5% absolute precision for a very low prevalence.
- Taking p from a very different population (for example, hospital data for a community study).
- Forgetting the design effect in multistage or cluster sampling.
- Applying the finite population correction when the population is large or unknown.
Key takeaways
- Expected prevalence and precision drive the sample size.
- Use relative or smaller absolute precision for low prevalences.
- Apply the finite population correction only for small, known populations.
- Multiply by the design effect for cluster sampling, then add a non-response allowance.
References
- Lwanga SK, Lemeshow S. Sample Size Determination in Health Studies: A Practical Manual. Geneva: World Health Organization; 1991.
- Charan J, Biswas T. How to calculate sample size for different study designs in medical research? Indian J Psychol Med. 2013;35(2):121–126. PubMed
- Dean AG, Sullivan KM, Soe MM. OpenEpi: Open Source Epidemiologic Statistics for Public Health. openepi.com
This article is for education. Always follow the requirements of your university, guide, ethics committee and target journal.