Adaptive Full Factorial Experiment — Adaptive Full Factorial Experimental Design
Adaptive Full Factorial Experimental Design · Also known as: adaptive full-factorial design, sequential full factorial experiment, adaptive complete factorial design, dynamic full factorial trial
An adaptive full factorial experiment is an experimental design that starts with a complete crossing of all factors and all their levels, then uses interim data to modify subsequent runs — dropping unpromising factor levels, adding new ones, or re-allocating replication — while preserving the full factorial structure within each phase. This integration of full factorial coverage with adaptive decision rules allows researchers to explore all main effects and interactions without committing to a fixed, inefficient run plan before any data are observed.
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 an adaptive full factorial experiment when you need complete information on all main effects and two-way interactions (making fractionation unacceptable) and you also have sufficient budget flexibility to adjust the design mid-study based on emerging results. It is well suited to process optimisation, engineering studies, and early-phase clinical or pharmaceutical research where the factor space is uncertain but the outcome is quantitative and measurable quickly. Do not use it when pre-specifying adaptation rules is impractical (e.g., very long outcome measurement windows that prevent timely interim analysis), when the number of factors is so large that a full factorial is infeasible even in one phase, or when regulatory requirements mandate a fully fixed design. If interaction information is not critical, a fractional factorial or response-surface design with adaptive augmentation is more efficient.
Strengths & limitations
- Preserves full factorial coverage within each phase, so all main effects and interactions remain estimable without confounding.
- Adaptive re-allocation of runs reduces wasted experimental effort on uninformative factor regions identified early in the study.
- Pre-registered adaptation rules maintain statistical validity and transparency, distinguishing this design from post-hoc data dredging.
- Combines the inferential completeness of a full factorial with the efficiency gains of sequential experimentation.
- Particularly powerful in process optimisation and engineering contexts where factor ranges are initially uncertain.
- Requires careful pre-specification of adaptation rules; poorly defined rules can introduce bias or inflate type-I error.
- More complex to plan, execute, and analyse than a fixed full factorial; demands statistical expertise in adaptive and sequential methods.
- Full factorial coverage per phase becomes logistically burdensome as the number of factors or levels grows.
- Interim analyses that trigger adaptations may require stopping the experiment early, which can reduce power if conservative boundaries are applied.
- Pooling data across phases with different factor-level configurations requires appropriate statistical modelling that is not always straightforward.
Frequently asked
How is this different from a standard adaptive experiment?
A standard adaptive experiment can use any run structure — including fractional, response-surface, or one-factor-at-a-time designs — and adapts based on interim data. An adaptive full factorial experiment specifically maintains complete factorial coverage within each phase, so every factor-level combination is observed. This rules out confounding of interactions but is more resource-intensive than adaptive designs that use fractional or sparse run structures.
Does adapting the design invalidate the statistical analysis?
Not if the adaptation rules were pre-specified before data collection began and if the analysis uses methods that account for the sequential, adaptive structure. Naive application of standard fixed-design ANOVA or regression to pooled adaptive data can inflate type-I error and bias estimates. Appropriate approaches include phase-stratified analysis, mixed models with phase as a blocking factor, and closed testing procedures that respect the adaptive nature of the design.
How many factors can I realistically include?
Full factorial coverage per phase grows exponentially with the number of factors and levels (a 2^k full factorial requires 2^k runs per phase). In practice, two to four factors at two levels (4–16 runs per phase) is feasible in most settings. Five or more factors at two levels (32+ runs) strains resources for most studies. If the factor count is large, consider a fractional factorial or response-surface design with adaptive augmentation rather than a full factorial approach.
Do I need ethics approval to adapt the design mid-study?
In human subjects research, yes — any planned adaptive design modification must be described in the original protocol and approved by the relevant ethics board. Adaptations should be pre-specified in the statistical analysis plan. Unplanned modifications after seeing the data require a protocol amendment and re-approval before being implemented.
Is this design suitable for very slow-outcome studies?
Generally no. Adaptive designs depend on timely interim data to guide adaptations. If the outcome takes months or years to measure, the experiment may be nearly complete before any interim results are available, negating the efficiency benefit of adaptability. For slow outcomes, consider a fully fixed full factorial design or a group-sequential design with very few pre-planned interim looks.
Sources
- Atkinson, A., Donev, A., & Tobias, R. (2007). Optimum Experimental Designs, with SAS. Oxford University Press. ISBN: 978-0199296606
- Montgomery, D. C. (2017). Design and Analysis of Experiments (9th ed.). Wiley. ISBN: 978-1119113478
How to cite this page
ScholarGate. (2026, June 3). Adaptive Full Factorial Experimental Design. ScholarGate. https://scholargate.app/en/experimental-design/adaptive-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.
- Adaptive ExperimentExperimental design↔ compare
- Factorial Randomized Controlled TrialExperimental design↔ compare
- Fractional Factorial ExperimentExperimental design↔ compare
- Full Factorial ExperimentExperimental design↔ compare
- Response Surface MethodologyExperimental design↔ compare