Model selection doesn't make sense to me

Help with analyzing a design and interpreting the results.
READ ONLY! If you have a question or comment, please e-mail stathelp@statease.com.

Moderators: Wayne, ShariK

Model selection doesn't make sense to me

Postby hoohaahii » Tue Feb 22, 2011 1:02 pm

Hello,

I inputted a performed experiment results in a Historical Data RSM form, and in the Fit Summary analysis "attached" all the numbers "compared to my knowledge in theory" imply that the quadratic model should be selected, but Design Expert still suggests a 2FI Model!

the Fractional design space graph though shows that the standard error of the design is somewhat lower for the 2FI? but how come that doesn't reflect in the Fit Summary results?

Can anyone please enlighten me on this issue?

Thank you very much,
A.Y.A
Attachments
Historical Data 18Feb1500.dxp
my design expert file
(7.72 KiB) Downloaded 166 times
hoohaahii
Registered Member
 
Posts: 4
Joined: Thu Feb 17, 2011 11:46 am

Re: Model selection doesn't make sense to me

Postby Wayne » Tue Feb 22, 2011 4:35 pm

The quadratic model makes sense to me too. The 2FI model is suggested because the full Quadratic model is slightly non-significant at the 0.05 level. The 2FI model only slightly beats it out. The fit summary only checks the full order models to suggest a starting point. If there are many insignificant terms in the higher order then that full model may be insignificant. In this case half of the quadratic terms are insignificant (there are only two). The insignificant B^2 term could be removed from the model using backward elimination. Once B^2 is removed the reduced quadratic model becomes significant in the ANOVA.

My rule of thumb is to carefully look over the fit summary. If the adjusted R-squared is higher for the higher order model, start there for model reduction, even if the software doesn't suggest anything.
Wayne
Stat-Ease Consultant
 
Posts: 314
Joined: Tue May 27, 2008 8:31 am

Re: Model selection doesn't make sense to me

Postby hoohaahii » Wed Feb 23, 2011 8:29 am

Thank you very much Mr. Wayne for your help, your answer cleared out a lot of thing for me :D I have a better command on my model now. :D
hoohaahii
Registered Member
 
Posts: 4
Joined: Thu Feb 17, 2011 11:46 am


Return to Analysis

Who is online

Users browsing this forum: No registered users and 1 guest

cron