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; } } }

A way to automatically trim

Jed_Campbell

What inspired this Wish List Request?

In both doing demos and watching others do demos where terms are removed one at a time from a model, but still satisfying model hierarchy, it's painfully slow. An example of this is in the excellent video for the DOE course lesson 4, the video on "Predicting Process Capability" where @monica_beals does this process three times (well, 2.5 times, as she takes a shortcut the last time). It seems like this should be faster/easier/simpler, especially if it's something that we all tend to show as a best practice.

Please describe the new or improved feature or solution you would like to see.

What I'd like to see is an "auto trim" (or perhaps a better name) link/button to automatically remove terms above a specified p-value, one at a time, while respecting model hierarchy. Basically, just automate the process that Monica does in the video linked above.

What would the value be to you and/or other users if this new feature is provided?

This would save time for users and also make demos less awkward. It's seriously slow doing this in front of potential customers.

Is this an external customer request?

No.

Is there a competitive software package that has this feature or solution that we should be aware of?

This is currently possible using the Stepwise platform in JMP, but that requires extra steps/explanations.

Will this impact a renewal, expansion, or adoption at a customer, and does this put an account at risk if we do not address this?

Not necessarily, but it would make JMP look easier and faster.

Status: Not Planned For Now

Comments

SamGardner

@Jed_Campbell what about using Backward stepwise regression? Either in GenReg or the Stepwise Platform. Also, the Fit Two-Level Screening platform does a good job of this. Easy DOE will do this also.

For well designed experiments, where the correlation of the parameter estimates is small, then selecting multiple factors at once from effects summary and then removing them will work well also and has almost no impact on the parameter estimates.

Jed_Campbell

Thanks@SamGardner .Yes, those are all good options. My suggestion is really geared more towards an awkwardness when we demo the software to new audiences. Usually, potential users aren't ready for stepwise or other platforms, and--especially when using a historic dataset (like tablet or powder metallurgy)--there are a lot of potentially insignificant factors to remove. Having the automatic button in this location would simply add a lot of "wow" factor to the demo. Or, in short, the suggestion is more of a marketing suggestion, and not so much a technical suggestion, if that makes sense.

Sarah-Sylvestre

mia_stephens

Hi @Jed_Campbell , given the discussion below, and the ability to automate model reduction using the methods described, we're marking this Not Planned for Now. I think there are also opinions around the risks of having an "easy" button for model reduction - i.e. that this oversimplification of finding the "best" model is generally not a good idea.