Randomization
Random run order — prevents time-related drift from being confounded with factor effects.
Also known as: run-order randomization, randomised order
Randomization requires the run order to be assigned at random — not in the design's standard-table order, not sorted by factor values. Only then can the influence of unobserved time-varying disturbances (tool wear, temperature drift, operator rotation) be separated from true factor effects.
- Complete randomization — every run number randomly assigned; ideal when setup times are non-critical
- Blocked randomization — randomize within homogeneous blocks (shift, batch, day)
- Split-plot — when a factor is hard to change (oven temperature), it is randomized in whole plots; the fast-changing factors are randomized within
See also
Used in
In the Algorithm Lab
Sources
- Montgomery, D. C. — Design and Analysis of Experiments, 10th Edition, Chapter 1
- Kleppmann, W. — Versuchsplanung, 10. Auflage, Hanser