topic Re: Using PRESS criteria to build multilinear models in Discussions

Using PRESS criteria to build multilinear models

dp30 — Tue, 20 Jul 2010 08:56:17 GMT

Hello,
When using the Stepwise platform, JMP helps you to make the model with the best R² / R²adj.
I would like to get the model with the best PRESS criteria (Predicted Error Sum of Squares). How would it be possible to do that?
I know you can get the PRESS information once the model is built, but I would like to be sure I am building the model with the best PRESS as possible.
Thank you for your help.

Re: Using PRESS criteria to build multilinear models

Tue, 20 Jul 2010 22:35:43 GMT

Since it's not available in stepwise I guess you are going to have to just build multiple models and compare the PRESS statistic. Maybe use stepwise first to get a short-list of potential models then compare the models in Fit Model. It would seem like a nice enhancement request to have it included in stepwise.

Re: Using PRESS criteria to build multilinear models

dp30 — Thu, 22 Jul 2010 06:44:08 GMT

Thank you for your response. I can't understand why neither JMP nor most popular statistical softwares can do this...