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; } } }

Help with DOE analysis

JRen

I have a DOE setup and need some help with analysis.

I have Three Factors

Z1-7 levels 6,7,7.5,8,8.5,9,10

Z2-7Levels 2,3,3.5,4,4.5,5,6

CR-3 levels 0.5,1.0,1.2

Three responses

P,V,T,and I added a sum of the three as I want to predict the lowest possible sum and what conditions to use to get it.

I will run the reccomended 21runs plus 4 extreme requested by the boss to stress the system.

What I would like is a prediction of the best combination of the different Factors to obtain the lowest sum response. Once I obtain data. I am not sure

how to process the date using jmp. Previously I would manually do it in excel and just pick the best graphs. Not efficient.

Thanks in advance.

Edge Response DOE r2.jmp

Accepted answers

Phil_Kay

Okay, that makes more sense and is a much more simple question to answer.

Yes, you are correct in your interpretation of the prediction profiler for your artificial data example.

As a general approach you can use the desirability functions in the profiler to find the most desirable combination of settings.

All comments

Phil_Kay

Hi,

That is really a big question. I don't think it is feasible to give you a comprehensive answer on how to analyse a DoE in a discussion thread.

I would point out that you have the model dialogue as a script to run in your data table. All you need to do is tell it the response column and run it. To understand how to interpret the report requires some learning though. You can refer to the help documentation, particularly the book on DoE. There is an example of running an analysis using one of the sample data files.

I will also make a shameless plug for my blog series on Why DoE.

Regards,
Phil

JRen

think I did not explain what I was looking for correctly.

I'm aware if I use the Model script it will drop the pvalues and show me which factor has the strongest correlation to my response. In the effect summary it will show the Pvalue and I theorize Zone 1 will be the largest contributor because that is just the physics of my equipment.

What I would like to do is determine which combination will give me the lowest sum peak. Since I am not doing a full factorial for every possibly combo.(I can change values in 0.1 increments) I would like predict what values to enter to get the lowest sum peak. Perhaps the top 4 conditions and then test them to verify since real world does not always follow calculations.

I put together some false data but it represents another experiment I think is similar to what I expect to see from this test.

So I think my question is really only related to prediction profiler. In my given example I run the script, set emphasis to effect screening and run. then I get the prediciton profiler. In this data set to me it seems that lowest peak should be Z1 at 10, Z2 at 6 and CR at 0.5 Am I reading this correctly?

Practice.jmp

Phil_Kay

Okay, that makes more sense and is a much more simple question to answer.

Yes, you are correct in your interpretation of the prediction profiler for your artificial data example.

As a general approach you can use the desirability functions in the profiler to find the most desirable combination of settings.

JRen

Great thanks. was able to find several vids explaining and showing how to play with desirablilty funcions. I'll run this experiment next week and plug the data and see how it goes.