Pilot Full Factorial Experiment
Also known as: pilot factorial design, pilot 2^k design, pilot complete factorial experiment, screening factorial pilot
A pilot full factorial experiment is a small-scale, complete crossing of all selected factors at all their levels, run before a definitive study to gather preliminary effect estimates, assess variability, and verify experimental logistics. It retains the complete combinatorial structure of a full factorial design — every combination of factor levels is tested — but is intentionally limited in scope (fewer replicates, narrower factor ranges) to conserve resources while maximising learning about factor effects and interactions before committing to a larger investigation.
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 pilot full factorial experiment when you are entering a novel experimental domain and need simultaneous estimates of multiple main effects and their interactions before committing to a large, expensive, or irreversible definitive experiment. It is especially valuable when the number of candidate factors is moderate (two to five), when factor interactions are plausible, and when little prior variance information is available for power calculations. Do NOT use it as a stand-alone confirmatory test — the pilot's small sample means estimates are imprecise and statistical conclusions are provisional. Also avoid it when the number of factors is large (six or more) and budget is tight; a fractional factorial or Plackett-Burman design will screen more factors with fewer runs.
Strengths & limitations
- Tests every factor combination, revealing interaction effects that one-factor-at-a-time pilots completely miss.
- Provides unconfounded estimates of all main effects and two-way interactions (for two-level designs).
- Generates empirical variance estimates essential for properly powering the definitive experiment.
- Exposes practical and logistical problems at low cost before full-scale commitment.
- Randomisation and blocking within the pilot protect internal validity even at small scale.
- Number of runs grows exponentially with factors (2^k); with k = 5 factors and 2 replicates, 64 runs are needed — often prohibitive as a pilot.
- Statistical power is low at typical pilot sample sizes; main effects can be detected in direction but precise magnitude estimates require the main study.
- A pilot with only one replicate per cell cannot estimate pure experimental error independently; curvature or active interactions may be confounded with error.
- Results are specific to the factor ranges tested; effects may not generalise outside the pilot's operating window.
Frequently asked
How is a pilot full factorial experiment different from a fractional factorial?
A full factorial tests every combination of factor levels, so all main effects and interactions are estimable without confounding. A fractional factorial tests only a selected subset of combinations — it is more resource-efficient but aliases (confounds) certain higher-order interactions with main effects or lower-order interactions. For a pilot with few factors (two to four), the full factorial is often affordable and preferred because it keeps all effects clear; with five or more factors the fractional factorial becomes necessary.
How many replicates do I need in a pilot?
One replicate per cell is the practical minimum; it supports effect estimation but not an independent estimate of experimental error. Adding two to four centre-point runs (for continuous factors) provides an error estimate and tests for curvature without doubling the run count. Two full replicates per cell is ideal when the budget allows, as it yields an honest error estimate and substantially improves effect detection.
Can I use the pilot data in the final analysis?
Yes, provided the pilot and main study used identical protocols, randomisation procedures, and measurement systems, and provided this was pre-planned. Pooling pilot and main-study data is a form of sequential experimentation. If any protocol change was made between phases, separate analyses are safer; using pilot data as a training set and the main-study data as a confirmatory set is a transparent alternative.
What response variable should I use in a pilot?
Use the same primary response as the planned main study. If the main study is expected to use a composite outcome, pilot a single well-measured component that is the most scientifically important. Variance estimates from the pilot response directly feed the power analysis for the main experiment, so consistency between pilot and main response is essential.
Is a pilot full factorial ethical and cost-justified?
In most applied and industrial research contexts a small pilot that prevents a large failed experiment is both ethically sound and cost-efficient. In clinical or animal research, the ethical calculus requires that the pilot expose the minimum number of subjects necessary to achieve the learning objective — a 2^2 or 2^3 design with one replicate often meets this criterion. Regulatory agencies and ethics boards generally expect pilot evidence of feasibility before large-scale human trials.
Sources
- Montgomery, D. C. (2017). Design and Analysis of Experiments (9th ed.). Wiley. ISBN: 978-1119113478
- Box, G. E. P., Hunter, J. S., & Hunter, W. G. (2005). Statistics for Experimenters: Design, Innovation, and Discovery (2nd ed.). Wiley. ISBN: 978-0471718130
How to cite this page
ScholarGate. (2026, June 3). Pilot Full Factorial Experiment. ScholarGate. https://scholargate.app/en/experimental-design/pilot-full-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.
- Fractional Factorial ExperimentExperimental design↔ compare
- Full Factorial ExperimentExperimental design↔ compare
- Response Surface MethodologyExperimental design↔ compare