When you need to test many variables but cannot afford the time or cost of testing every possible combination, use this method. It selectively chooses a subset of experiments, allowing you to identify which factors are most important without running every single scenario, making it ideal for initial screening phases.
Efficiency through smart selection
A fractional factorial design works by systematically selecting a subset of experiments from a larger set. Instead of testing all possible combinations of your factors, this method uses a structured mathematical framework to ensure that the most important information remains preserved.
By focusing on the primary effects of your variables, you can extract meaningful insights even when testing 5 to 8 factors at once. This approach prevents the need for an unmanageable number of runs, ensuring your project remains within scope.
Understanding design resolution
Lattice uses resolution levels to help you balance the number of experiments against the clarity of your results. Resolution III is the most aggressive option, useful for quick screening when you only need to identify major drivers. Resolution IV is the standard, keeping main effects clean, while Resolution V is the most detailed, providing results close to a full design.
Choosing the right resolution allows you to tailor the experiment to your specific constraints. If your goal is simply to prune a long list of potential factors, a lower resolution is often sufficient and highly efficient.
Deterministic and repeatable setup
Because this design is based on fixed mathematical principles, the output is entirely deterministic. When you provide your factors and desired resolution, Lattice generates a specific, reliable matrix that you can replicate or audit at any time.
The process includes built-in randomization of the execution order. This helps ensure that external variables—such as device drift or changes over time—do not bias your results, giving you higher confidence in the final conclusions you draw from your data.
1 · Intent → method
An LLM picks doe_generate_fractional_factorial from a fixed catalog.
2 · Method → numbers
Deterministic Python engine runs the math. Same input → same output.
3 · Numbers → plain language
A second LLM translates the result into your domain’s vocabulary.
Why should I use a fraction instead of all combinations?
Testing every possible combination (a full factorial) grows exponentially as you add factors. This method allows you to identify which variables matter most by running only a carefully selected portion of those tests, significantly reducing your laboratory or compute time.
What does 'Resolution' mean in this design?
Resolution describes how clearly the method separates the effects of your variables. A higher resolution provides a cleaner look at your results but requires more runs, while a lower resolution is more efficient at screening when you have many factors to investigate.
Tool input schema
Schema for doe_generate_fractional_factorial not exported yet (run pnpm export:registry).