cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
  • Image to Table JMP extension imports single or batch images as data tables. Useful to extract curves from pictures or convert image data to curve data to use in analyses like Functional Data Explorer. Download and install the extension.

Discussions

Solve problems, and share tips and tricks with other JMP users.
Choose Language Hide Translation Bar
carlosariasq
Level II

Full Factorial with center points

Dear community, good day 

 

I have a small question. I have a full factorial DOE with three factors (2 continuous X1 and X2 and 1 categorical X3). 

At the moment of fitting the response, only one of the continuous factor (X1) is active, but this same factor (X1) suggests non-linear effects given the center points. I dropped the other continuous factor (X2) and I have a set of triplicates of the center points (another block because they were performed in another days).

Including a quadratic effect for this active continuous factor (X1), renders my categorical factor (X3) active, the quadratic effect active (X1xX1) and the interaction of categorical x continuous active (X1xX3). Despite this providing a better and more accurate model, I was wondering if this is actually possible to do, since I know a full factorial cannot fit a higher order model 

 

Thanks for your feedback 

5 REPLIES 5
MathStatChem
Level VII

Re: Full Factorial with center points

check out this previous discussion:  https://community.jmp.com/t5/Discussions/Full-Factorial-Center-Points/m-p/58736

short answer is no, you can't estimate the quadratic effect independently, but you can do a statistical test to detect curvature, which would let you know that there is possibly one or more quadratic. effects.  An augmentation of the design, running a small number of additional experiments, would allow you to estimate the quadratic effects.  

carlosariasq
Level II

Re: Full Factorial with center points

Thanks a lot for your reply.

One last question: What would suggest the inclusion of this quadratic effect (despite incorrect) improving the Lack of Fit?

Thanks a lot

Victor_G
Super User

Re: Full Factorial with center points

Hi @carlosariasq,

Just to explain the response from @MathStatChem and answer your question:

Centre points are used to determine if there might be any curvature effect on the response(s). To determine which factor(s) is responsible for this curvature effect, it would require the use of middle levels for the different continuous factors. You can still use centre points to estimate one quadratic effect, but if you have several continuous factors, you may not know with confidence which factor(s) may be responsible of the curvature effect in the response(s). The improvement on the lack-of-fit you see by adding one quadratic effect could imply that you may have to search for the quadratic effects of the continuous factors. Note that you can also look at the different residual plots, to see if a non-random/curved pattern may show a missing term in the model.

In order to differentiate and determine from which factor(s) come the curvature effect on the response, you may need to Augment your design and add quadratic effects in the model for the continuous factors.  

Hope this complementary answer can help you,

Victor GUILLER

"It is not unusual for a well-designed experiment to analyze itself" (Box, Hunter and Hunter)
statman
Super User

Re: Full Factorial with center points

Both Mathstatchem and Victor gave appropriate answers. I want to ask, how did you create a center point for the categorical X? There is no center point. Did you just choose 1 of the 2 levels and center points for the continuous factors? For center points, X1*X1=X2*X2=X3*X3. Since these effects are confounded, there is no exact way to determine if it is one or more of those effects that are active. Victor's suggestion of looking at residuals for curvature is a great idea, but the only true way is to run additional treatments.

BTW, multiple replicates of the center point randomly run throughout the experiment is a great way to evaluate stability and estimate MSE for statistical tests.

"All models are wrong, some are useful" G.E.P. Box
carlosariasq
Level II

Re: Full Factorial with center points

Thank you very much for your feedback. Concerning the question, I read in Mee, 2009 an example in which the authors performed center points at the quantitative factors but varying the two levels of the only qualitative one. I did the same but including the center conditions of my quantitative factors.

Recommended Articles