Crossover Fractional Factorial Experiment
Crossover Fractional Factorial Experimental Design · Also known as: crossover FF design, within-subject fractional factorial, repeated-measures fractional factorial, crossover FFE
A crossover fractional factorial experiment is a within-subject design in which each participant receives a strategically chosen subset of all possible factor-level combinations in a defined sequence, with washout periods between treatment periods. By combining the run-economy of fractional factorial designs with the within-subject efficiency of crossover designs, it allows estimation of main effects and selected interactions while controlling for between-subject variability using far fewer participants and experimental runs than a full factorial crossover.
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 crossover fractional factorial experiment when: (1) each participant can feasibly receive multiple treatment combinations sequentially, (2) you need to screen or estimate effects of several factors simultaneously but cannot afford a full factorial crossover, (3) stable carryover-free conditions can be achieved through washout, and (4) within-subject variability is large relative to between-subject variability so that individual-as-own-control control is highly beneficial. Ideal contexts include early-phase clinical pharmacology, sensory evaluation, human factors research, and agricultural or agronomy trials with plot effects. Do NOT use this design when: carryover effects cannot be eliminated (irreversible treatments, learning effects that persist), the number of required periods per subject would be ethically or practically unacceptable, or when treatment-by-period interactions are expected to be strong.
Strengths & limitations
- Combines two efficiency gains simultaneously: fractional factorial reduces the number of treatment combinations, and crossover removes between-subject noise.
- Requires substantially fewer participants than a parallel-group full factorial design to achieve equivalent statistical power for main effects.
- Allows estimation of main effects and selected two-factor interactions from a single well-structured study.
- Well-suited to situations where participant recruitment is costly or sample sizes are inherently limited.
- Systematic design tables (resolution III/IV/V fractions, Williams squares) are available, making implementation straightforward.
- Confounded effects: high-order interactions are deliberately aliased; their separate estimation requires additional runs.
- Carryover is a critical threat; if washout is insufficient, period-by-treatment effects bias estimates and the crossover advantage is lost.
- Requires a relatively long study duration per participant, increasing dropout risk and the possibility of time-varying confounders.
- Analysis is more complex than a simple crossover or a simple factorial; mixed-effects modeling expertise is needed.
- Assumes that the participant population is stable across all periods (no disease progression or learning effects that interact with treatment).
Frequently asked
How do I choose the fraction resolution for a crossover fractional factorial?
Resolution IV is usually the minimum acceptable choice for applied research: it ensures no main effect is aliased with any other main effect or two-factor interaction. Resolution III fractions are risky unless prior theory rules out all two-factor interactions. Resolution V fractions preserve two-factor interaction estimability but require more periods per participant.
How long should the washout period be?
Washout should be at least five half-lives of the treatment effect for pharmacological interventions. For behavioral or educational treatments, washout is more difficult to define; if persistent learning or irreversible change is expected, a crossover design is inappropriate regardless of washout length.
Can I use a crossover fractional factorial with more than two levels per factor?
Yes, but design construction becomes more complex. Mixed-level fractional factorial crossovers (some factors at 2 levels, others at 3) require specialist design tables or algorithmic search (e.g., D-optimal designs). Standard 2^(k-p) tables apply only to two-level factors.
What statistical model should I use for analysis?
A linear mixed-effects model (or generalized linear mixed model for non-normal outcomes) with fixed effects for treatment, period, and sequence, and a random intercept for subject. Check for carryover by including a first-order carryover term; if significant, interpret treatment estimates cautiously and consider a sensitivity analysis excluding the first period.
How does this differ from a blocked fractional factorial experiment?
In a blocked fractional factorial, experimental units are grouped into blocks to control a nuisance variable, but each unit typically receives only one treatment. In a crossover fractional factorial, the same participant serves as their own control across multiple treatment periods, providing much stronger within-subject control — but also introducing the carryover risk that blocking designs avoid.
Sources
- Senn, S. (2002). Cross-over Trials in Clinical Research (2nd ed.). Wiley. ISBN: 978-0471496533
- 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). Crossover Fractional Factorial Experimental Design. ScholarGate. https://scholargate.app/en/experimental-design/crossover-fractional-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.
- Crossover Full Factorial ExperimentExperimental design↔ compare
- Crossover Randomized Controlled TrialExperimental design↔ compare
- Factorial ExperimentExperimental design↔ compare
- Fractional Factorial ExperimentExperimental design↔ compare
- Full Factorial ExperimentExperimental design↔ compare