Power is 1 minus beta, where beta is the Type II error rate, the chance of missing an effect that is really there. It rises with sample size, effect size and alpha, and falls as noise grows. Studies are usually sized for 80 or 90 percent power to detect the smallest difference that matters; detecting a 1 sigma difference between two means at alpha 0.05 with 90 percent power takes about 22 units per group.
Four wafers per leg gives you maybe 30 percent power for a 1 nm shift. Either add wafers or don't bother running it.
A non-significant result from a low-power study means the test could not tell, not that there is no effect. Size the experiment before running it.
Also heard as
- statistical power
Related terms
p-value
Probability of seeing a result at least this extreme if there were truly no effect, used to judge statistical significance.
Significant
Describes a result unlikely to have come from chance alone at a stated risk level, which is not the same as important.
t-test
Test that compares a mean against a target, or two means against each other, using the t distribution.
Replicate
Independent rerun of the same factor settings from scratch, which gives a true estimate of experimental error.