Profiler display of multilevel categorical Factors in “Fit Group”

[#1]

Hello,

I search a description what are the advantages in detail of the underneath shown Profiler display of multilevel categorical Factors in “Fit Group” with "personallity Stepwise". (two Top left graphs vs six left graphs in the lower Picture) Is there a paper available that explains why the two personallity/methods using different outputs. Can I switch the “Personallity: Stepwise” to the Graphical display used in “Standard least Squares”   (upper Graph), what I like currently much more.

I know, I can go via “Model Dialog” and doctoring the Factors to get the display I currently want to have. But I have multi Y’s and that would cause a mess of manual work. Hence I’m looking for a stepwise method with least squares Display.

A script would be helpful if it is easy to apply. I’m beginner with Scripting.

Thanks Peter

[#2]

Stepwise models containing categorical variables sometimes result in the creation of dummy variables.  I think it's the case where the model contains two factor interactions.

That doesn't happen with standard least squares.  

As you say, you can go back to the model dialog and edit the model specification, but it's not ideal.  

I don't really want to see the dummy variables, and certainly not when I push the model back into a least squares model -  I would have thought JMP could fix this, although maybe I am missing some mechanism that already exists?

[#3]

In the Stepwise platform, under rules it likely says "Combine". Try changing it to "Whole Effects". That should avoid some of those "strange" terms that get added to the model. It still may not be suitable for your needs, but it is a good place to start.

[#4]

To follow up on @Dan_Obermiller's suggestion, you won't see a difference in the Stepwise term list when you change the rule to "Whole Effects." You will see the difference when you click Make Model or Run Model.

[#5]

Thanks Dan, as you wrote, it mitigates my problem. But still plenty of  these Terms remain in my Model.

-Peter-