Advanced Cluster Sample Size Calculator

Calculator Inputs

Study mode

Choose the target parameter for the survey.

Confidence level

Higher confidence increases the required sample.

Margin of error

Use percent points for proportions or absolute units for means.

Expected proportion (%)

Typical conservative choice is 50%.

Standard deviation

Needed only for mean-based estimation.

Average cluster size

Average sampled respondents per cluster.

Intracluster correlation (ICC)

Use a pilot estimate or published benchmark.

Design effect mode

Auto uses 1 + (m - 1) × ICC.

Manual design effect

Used when prior studies already report design effect.

Population size

Optional, but useful for finite populations.

Nonresponse rate (%)

Inflates planned interviews to protect precision.

Apply finite population correction

Example Data Table

Scenario	Mode	Key Inputs	Design Effect	Completed Target	Planned Interviews	Clusters
National household survey	Proportion	95%, p=50%, e=5%, m=20, ICC=0.02, N=10,000, nonresponse=10%	1.38	503.44	560	28
Community vaccination audit	Proportion	95%, p=30%, e=4%, m=15, ICC=0.01, nonresponse=15%	1.14	574.80	690	46
Mean score benchmark study	Mean	95%, SD=12, e=2, m=10, ICC=0.03, nonresponse=10%	1.27	175.64	200	20
School wellbeing survey	Proportion	99%, p=40%, e=4%, m=25, ICC=0.02, N=8,000, nonresponse=12%	1.48	828.93	950	38

Formula Used

The calculator begins with a simple-random-sample requirement, then inflates it for clustering, optionally applies finite population correction, and finally adjusts for nonresponse.

For proportions:
n_srs = (Z² × p × (1 - p)) / e²

For means:
n_srs = (Z² × σ²) / e²

Automatic design effect:
DEFF = 1 + (m - 1) × ICC

Cluster-adjusted completed sample:
n_cluster = n_srs × DEFF

Finite population correction:
n_fpc = n_cluster / [1 + ((n_cluster - 1) / N)]

Nonresponse adjustment:
n_planned = n_fpc / response_rate

Required clusters:
clusters = ceil(n_planned / m)

Where Z is the confidence multiplier, p is the expected proportion, σ is the standard deviation, e is the desired margin of error, m is average cluster size, and N is population size.

How to Use This Calculator

Select whether your study estimates a proportion or a mean.
Choose the confidence level and enter the margin of error you can tolerate.
Enter either the expected proportion or the standard deviation, depending on study mode.
Set the average cluster size and either provide ICC or a manual design effect.
Add population size when the target population is limited and enable finite population correction.
Enter the expected nonresponse rate to convert required completed cases into planned interviews.
Press calculate and review completed sample, planned interviews, cluster count, and the Plotly sensitivity chart.
Use the CSV or PDF buttons to export the result summary.

FAQs

1. What is a cluster sample size?

It is the number of interviews or observations needed when people are sampled in groups, such as schools, villages, or clinics. Because people within one cluster tend to be similar, the needed sample is usually larger than a simple random sample.

2. Why does ICC matter?

ICC measures how similar responses are inside the same cluster. A higher ICC means more overlap in information, which increases the design effect and raises the required sample size.

3. When should I use 50% as the proportion?

Use 50% when you do not have a reliable prior estimate. It gives the most conservative proportion-based sample size because p × (1 − p) is largest at 0.50.

4. What is design effect?

Design effect compares the variance under clustered sampling with the variance under simple random sampling. A value of 1 means no inflation. Larger values mean you need more completed interviews.

5. Should I apply finite population correction?

Apply it when the target population is not very large and your sample is a noticeable share of that population. It often reduces the required completed sample.

6. Why does the final interview count round upward?

Field plans usually need full clusters, not fractions of clusters. The calculator rounds up to the next whole cluster so your design can actually be implemented.

7. What is the difference between completed sample and planned interviews?

Completed sample is the number of usable responses needed for precision. Planned interviews are the larger fieldwork target after allowing for expected nonresponse and cluster rounding.

8. Can I use a manual design effect?

Yes. Use manual design effect when prior surveys, pilot studies, or technical guidelines already provide a defensible inflation factor. That can be more practical than estimating ICC directly.