DoE (Design of Experiments)

Systematic experiment planning so that a minimum number of runs yields maximum insight into factors, main effects, and interactions.

Also known as: experimental design, designed experiment, DoE

Design of Experiments (DoE) studies several inputs simultaneously under a fixed experimental plan. Compared to the classical "one factor at a time" approach: fewer runs, statistically anchored effects, and interactions are detected instead of missed.

  • Screening — which of the many candidate factors matter at all?
  • Optimisation — for the few important factors, find the best setting (Response Surface, CCD, Box-Behnken)
  • Robustness — settings where the response stays insensitive to noise factors (Taguchi, dispersion DoE)
  • Mixtures — recipes whose proportions sum to 100 % (Simplex-Lattice, Simplex-Centroid)

See also

Used in

In the Algorithm Lab

Sources

  • Montgomery, D. C. — Design and Analysis of Experiments, 10th Edition, Wiley
  • Box, G. E. P., Hunter, W. G., Hunter, J. S. — Statistics for Experimenters, 2nd Edition, Wiley
  • Kleppmann, W. — Versuchsplanung, 10. Auflage, Hanser