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; } } }

SVEM Prediction formula

frankderuyck

I made a SVEM analysis and would like to use the formula to comupute process control ranges using the Design Space Profiler. I saved the prediction formula and can't save a prediction formula with process parameters, only a Mean formula here below?

Is it possible to get an expression as a function of the process parameters that can be used in the design space profiler? Or can I use the Maximum Likelihood expression reported together with SVEM in the GENREG report?

Accepted answers

Victor_G

Hi @frankderuyck,

Yes it's correct, please read my previous answer. I explained why you get some Prediction sample formula in the Profiler : some of these Prediction Formula are constant values, so JMP can't reduce it anymore, so they are part of the Profiler. If you replace the Prediction sample formula which have constant values by the constant values directly in the Pred SVEM formula, you won't have them in the Profiler.

I already did this with your table and prepared a script in your dataset with the correct and expected Profiler.

All comments

Victor_G

Hi @frankderuyck,

When launching the Profiler, click on Expand Intermediate Formulas. This will indicate to JMP that instead of displaying your prediction formula depending on the other sample prediction formula, it has to develop the equation of each sample prediction formula to get to the initial factors.

Hope this answer will solve your problem,

MathStatChem

If you create the formula as a sum divided by the number of prediction formulas, you can use that as @Victor_G says, with the expanded formulas option, and then from the profiler you can save the expanded formula as a new column. You can also further simplify that formula (open the formula editor and select "Simplify" from the red triangle). That creates a a prediction formula with each model effect have a single coefficient.

MathStatChem

Note you have to use a sum using "+"'s , not the Sum( ) or Mean( ) function for this to work.

frankderuyck

I don't find the Expand Intermediate Formulas in the Profiler menu?

MathStatChem

It is an option on the launch dialog.

frankderuyck

Hi, thanks for your inputs!

I checked both (1) Expand Intermediate fromulas" but this gave the same result as above? Strange that when I click on relaunch the Expand option is not marked, so guess this did not work properly?

I also tried the the Simplify option by going to the Prediction Formula SVEM column, opening the formula editor and clicking on Simplify and the formula remains unchanged as shown above; going back to formula editor I see that Simplify is not marked so I guess this is not active?

I would like to use the SVEM model formula for process control range specification for it is a validated, more reliable & noise-robust model. I will now - with only one response Y - do this directly from the prediction profiler. With more responses Y I will need the profiler.

MathStatChem

Are you using Generalized Regression to develop the SVEM model? That platform allows you to save the simplified formula, in version 18 and 19, when you choose "Save Prediction Formula". What you are using are the resampled prediction formulas, I think.