Calculate design effect, average cluster size, or ICC from any two values using the formula DE = 1 + (m – 1) × ICC for clustered samples.
Design Effect Formula
This calculator estimates the clustering variance multiplier for equal-size clusters, or approximates it using average cluster size, with nonnegative ICC. Unequal cluster sizes, survey weights and other design components may require a different design effect. The formula used here is:
To solve for average cluster size, the formula is rearranged as:
To solve for the intraclass correlation coefficient, the formula is rearranged as:
- DE = design effect
- m = average cluster size
- ICC = intraclass correlation coefficient
Select design effect, average cluster size, or ICC as the quantity to solve for, then enter the other two values. ICC must be entered as a proportion from 0 to 1 in this planning model.
Typical Design Effect Interpretation
Within this model, a design effect of 2 gives twice the variance of a simple random sample of the same size, or an approximate effective sample size of half the nominal size. The categories below are illustrative, not universal thresholds.
| Design Effect | Interpretation | Effect on Precision |
|---|---|---|
| 1.00 | No clustering effect | Same as simple random sampling |
| 1.01 to 1.50 | Small clustering effect | Slight loss of precision |
| 1.51 to 2.50 | Moderate clustering effect | Noticeable increase in required sample size |
| Above 2.50 | Large clustering effect | Substantial loss of precision |
Common ICC Ranges for Clustered Data
Even a small positive ICC can create a large design effect with large clusters. The ranges below are illustrative descriptions; an appropriate ICC depends on the outcome, measurement and target population.
| ICC Value | Within-Cluster Similarity | Typical Meaning |
|---|---|---|
| 0.00 | None | People or observations within clusters are not more similar than random observations |
| 0.01 to 0.05 | Low | Common in many survey and public health settings |
| 0.06 to 0.15 | Moderate | Clusters are meaningfully similar internally |
| Above 0.15 | High | Cluster membership has a strong relationship with the outcome |
Example Calculations
Example 1: Calculate Design Effect
Suppose the average cluster size is 25 and the ICC is 0.04.
The design effect is 1.96. This means the clustered design has nearly twice the variance of a simple random sample of the same size.
Example 2: Calculate ICC
Suppose the design effect is 2.20 and the average cluster size is 31.
The intraclass correlation coefficient is 0.04.
FAQ
What does a design effect greater than 1 mean?
A design effect greater than 1 means the clustered sample is less statistically efficient than a simple random sample. Observations within the same cluster are more similar to each other, so each additional observation adds less independent information than it would in a simple random sample.
Can the design effect be equal to 1?
Yes. The design effect equals 1 when the ICC is 0 or when the average cluster size is 1. In that case, clustering does not increase the variance compared with simple random sampling.
Why does average cluster size matter?
Average cluster size matters because the effect of ICC grows as clusters get larger. With a larger cluster size, more observations are grouped with similar observations, so the amount of independent information increases more slowly.
