Cluster Randomized Factorial Experiment
Also known as: cluster-randomized factorial design, group-randomized factorial trial, CRT factorial, clustered factorial experiment
A cluster randomized factorial experiment assigns intact groups (clusters such as schools, clinics, or communities) at random to all combinations of two or more treatment factors, enabling simultaneous evaluation of multiple interventions and their interactions while respecting the natural grouping of participants. It merges the logistical and ethical advantages of cluster randomization with the efficiency of factorial design.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use a cluster randomized factorial experiment when: (1) the intervention is delivered at the group level and individual randomization is impractical or would cause contamination; (2) you want to test two or more interventions simultaneously and detect potential synergistic or antagonistic interactions; and (3) you have enough clusters to achieve adequate power after accounting for the ICC. It is especially valuable in public health, education, and community-based research where interventions are inherently group-level. Do NOT use it when you have too few clusters (fewer than approximately 4-6 per cell), when the ICC is very high and inflates required sample sizes prohibitively, when interventions cannot be delivered independently to clusters (i.e., spillover is unavoidable), or when a simple head-to-head comparison of a single intervention suffices.
Strengths & limitations
- Evaluates multiple interventions and their interaction in a single study, making efficient use of research resources.
- Cluster randomization protects against contamination between treatment conditions in settings where group-level delivery is unavoidable.
- The factorial structure preserves statistical power for main effects: each participant contributes data to estimating the effect of every factor.
- Interaction tests reveal whether combining interventions produces synergistic or antagonistic effects that single-arm trials cannot detect.
- Well-suited to real-world implementation settings (schools, clinics, communities) where policy-relevant combinations of interventions are of interest.
- Requires a larger number of clusters than a simple cluster RCT to power interaction tests adequately.
- Intraclass correlation reduces the effective sample size; underestimating the ICC leads to underpowered studies.
- Statistical analysis is more complex than individual randomization, requiring multilevel or GEE models.
- Assumes interventions can be delivered to clusters independently; if combined delivery fundamentally changes implementation fidelity, the factorial assumption is violated.
- Recruiting and managing many clusters with different treatment combinations increases logistical and coordination demands.
Frequently asked
How is this different from a standard cluster RCT?
A standard cluster RCT compares a single intervention against control by randomizing clusters. A cluster randomized factorial experiment randomizes clusters to all combinations of two or more factors, enabling estimation of each factor's main effect and their interaction within the same trial. The factorial structure is more efficient when all factors are of interest.
How do I calculate sample size?
Power calculations must account for the intraclass correlation coefficient (ICC), the average cluster size, the number of factorial cells, and — if interaction effects are of primary interest — additional multipliers for the interaction test. Standard software (e.g., Optimal Design, PowerUp!, or R packages such as clusterPower) supports these calculations. Underestimating the ICC is the most common error leading to underpowered studies.
What if the ICC is very high?
A high ICC means most variance lies between clusters rather than within them, which greatly reduces effective sample size. If the ICC is high (e.g., above 0.10 in education settings), the required number of clusters may become prohibitive. In such cases, consider a simpler design (single-factor cluster RCT) or explore whether individual randomization is feasible without contamination risk.
Do I need the same number of clusters in every cell?
Balanced designs (equal clusters per cell) are optimal for power and simplify analysis. Unbalanced designs are permissible but reduce power and require more careful mixed-model specification. Stratified randomization helps achieve approximate balance across covariates and cells.
Can I use this design for more than two factors?
Yes — three or more factors can be crossed in a cluster randomized factorial experiment, but the number of treatment cells grows multiplicatively (e.g., 2x2x2 yields eight cells). Each cell must contain enough clusters for adequate power, so three-factor factorial cluster RCTs require very large numbers of clusters and are rare in practice outside of large multisite trials.
Sources
- Murray, D. M. (1998). Design and Analysis of Group-Randomized Trials. Oxford University Press. ISBN: 978-0195120912
- Donner, A., & Klar, N. (2000). Design and Analysis of Cluster Randomization Trials in Health Research. Arnold. ISBN: 978-0340691533
How to cite this page
ScholarGate. (2026, June 3). Cluster Randomized Factorial Experiment. ScholarGate. https://scholargate.app/en/experimental-design/cluster-randomized-factorial-experiment
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Cluster Randomized Controlled TrialExperimental design↔ compare
- Factorial ExperimentExperimental design↔ compare
- Factorial Randomized Controlled TrialExperimental design↔ compare
- Fractional Factorial ExperimentExperimental design↔ compare
- Multi-arm experimentExperimental design↔ compare