A robust process holds its output steady when noise factors move: ambient humidity, incoming material variation, tool-to-tool differences, aging. Robust design, as in Taguchi parameter design, varies noise factors on purpose in the experiment and picks control factor settings that flatten the response to them, exploiting interactions between control and noise factors. In data analysis, a robust method is one that resists outliers, such as using the median instead of the mean.
Setting B is slightly lower on average, but it's robust to incoming film thickness. We'll take it.
Also heard as
- robustness
Related terms
Interaction
Situation where the effect of one factor depends on the level of another, so their combined effect is not simply additive.
Sweet spot
Region of factor settings where every response meets its target at once, found by overlaying response contour plots.
DOE
Design of experiments: planning a set of runs that vary several factors at once so their effects can be separated.
Outlier
Observation that sits far from the pattern of the rest of the data, either a genuine extreme or an error.