Full Versus Fractional Factorial Designs

Evaluate tradeoffs between full factor evaluation and fractional screening matrices using resolution indices, generator definitions, and confounding patterns.

2026-10-27
Arthur Penn
4 min read
Design Selection4 min read
Published: 2026-10-27
Full Versus Fractional Factorial Designs

Contents Overview

Article Sections

Full Factorial: Every Combination

A two-level full factorial with k factors contains 2k treatment combinations. Two factors need 4 runs, three need 8, five need 32, and seven need 128 before any replication.

The advantage is completeness: all main effects and all interactions can be estimated separately. The cost grows quickly as factors are added.

Key Principle

The effect of one factor may depend on the level of another.


Fractional Factorial: A Planned Subset

A fractional factorial, written 2k-p, runs a balanced fraction of the full set. A 25-1 design studies five factors in 16 runs instead of 32.

The fraction is defined by generators, for example setting the column for factor E equal to the product ABCD. That choice determines which effects share the same column and therefore cannot be told apart.


Aliasing and Design Resolution

Effects that share a column are aliased. Resolution summarises the pattern: in a Resolution III design main effects are aliased with two-factor interactions; in Resolution IV main effects are clear of two-factor interactions, but two-factor interactions are aliased with each other; in Resolution V main effects and two-factor interactions are clear of each other.

A practical rule is to use fractional designs for screening many candidate factors, then follow up with a full factorial or a higher-resolution fraction on the few factors that matter.

Swipe to view all columns

DesignFactorsRunsTypical use
22 full24Teaching, small studies
23 full38Complete interaction study
25-1 fraction516Screening with Resolution V
27-4 fraction78Early screening, Resolution III

Key Takeaways

  • Full factorials estimate everything but grow as 2k.
  • Fractional designs trade information for runs through aliasing.
  • Check the alias structure before interpreting any effect.
Arthur Penn

Arthur Penn

Contributor

Educational article for FactorInteraction Notebook. It explains planning logic and does not report experimental results or statistical significance.

Editorial Method