nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Augment Design: Enter Numbers of Runs

Peng66

I am performing materials design with DOE. I already have 150 experimental data. There are more 15 variables. Now I want to add additional runs by Augment Design. I am confused how to determine the Numbers of Runs. How do I know the new experiments from augment design is good or not?

PS: I have tried the Numbers of Runs from 300 to 100000, the D-eff. is still low and never reach 100. it can reach 40 at round 8000, but 8000 new experiments are too much. What should I do ?

Accepted answers

All comments

Phil_Kay

D-eff might not be the most useful design diagnostic for you to use in this case. My guess is that the existing 150 runs have some strong correlation between effects so even when you augment with a lot of new runs the complete design is far from orthogonal.

You should consider the power to detect effects. You can probably get a good estimate of experimental "noise" from your existing 150 runs. Then you just need to decide what size of effect it is important for you to detect.

I would recommend that you read Optimal Design of Experiments: a case study approach by Goos and Jones as this goes into a lot of detail about many of the design diagnostics.

Peng66

Thank you very much. phil_kay.

I have read the first 3 chapter yet and get a lot of details. However, I am confused what should I do in real case since the book did not show how to do it.

Could you tell me how to consider the power to detect the effect. A tutorial or example will be the best.

Phil_Kay

The Mastering JMP webcasts are a good place to see case study examples using JMP. There is a recorded webcast about design diagnostics that should help.

Peng66

Thank you very much