Sample size determination and effect size

DOEs are a very complex topic that is difficult to cover in a post, but I will try to cover the basics.

• Starting with your use of the word repetitions, DOE distinguishes between repeats and replicates.
  • Repeats are multiple measurements taken during a single experimental run and only include short-term variation. They provide minimal value to the DOE.
  • Replicates are different from repeats by performing a specific experimental run multiple times. This includes setup variation and provides additional degrees of freedom to the experiment. Replicates add value to the DOE, but also add cost by increasing the size of the experiment.
• Regarding the process variance, do you have any historical data of any type? This could be from a capability study or even production records. It does not need to be precise; a ballpark estimate is sufficient. You can also run the calculation for various estimates to see how sensitive the number is to the variation.
• Regarding the effect size, you should already know how much the process swings high to low. Pick an effect size larger than this that is large enough to justify the expense of running this experiment. Don't worry about the multiple factors. ANOVA looks for the largest effect.

 

I do not recommend that. It would be simpler to monitor your process for a day, collecting the data needed to calculate the variance.

 

Miner is correct. Assessing power after the effect is a mathematical exercise that has no meaning.