A replicate repeats a treatment combination with a full new setup: new wafer, fresh recipe load, separate run. It captures all the run-to-run sources of noise, so it gives a pure error estimate for significance tests. Measuring the same wafer three times is a repeat, not a replicate, and treating repeats as replicates makes the error look small and almost every effect look significant.
Those three readings are repeats on one wafer, not replicates. Your error bar is gauge noise, not process noise.
Ask whether the setup was torn down between runs. The answer decides whether you have replicates or repeats.
Also heard as
- replication
- true replicate
Related terms
Randomization
Running experimental trials in a random order so unknown time trends and nuisance effects do not bias the results.
Center point
Run with every numeric factor set halfway between its low and high levels, used to detect curvature and estimate error.
Factorial design
Experiment that runs every combination of the chosen factor levels, so each main effect and interaction can be estimated.
Repeatability
Scatter in the readings when one person and one gauge measure the same part over and over with nothing else changed.