cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Try the Materials Informatics Toolkit, which is designed to easily handle SMILES data. This and other helpful add-ins are available in the JMP® Marketplace
Choose Language Hide Translation Bar
DHBWDozent
Level II

Factor Constraints disappears center points and some other strange behavior

Hello.

I tried to create a custom design with 5 continous, 2 discrete numeric and 1 categorial factor (all 2 levels). I wanted 4 center points and 3 replicates. This works fine, until I try to define a factor constraint. I have a continues factor "storage period" with the level 0 hours and 48 hours and 2 storage places (categorial), 1 in the production environment, 1 in a controlled climate room. At the storage period 0 hours, the parts have no choice, the have to be stored all in the climate room. If I now try to constrain  the factor storage period 0 with the storage place "climate room" (also cannot be stored in the environment) and make the same design as before, all center points disappear.

 

Why? I cannot understand why this happens. I cannot see any reason, why with this constraint it would be not possible to set center points.

 

Another strange behavior, if I set the factor constrains as mentioned:

 

DHBWDozent_0-1626512948664.png

DHBWDozent_1-1626513537662.png

 

The design in the column "Lagerdauer" (storage period) shows values 2,4. Why, I never defined something like a storage period of 2,4 hours? Should it be 24, in the center of "Lagerdauer", or what it the reason of this 2,4?

DHBWDozent_2-1626513605425.png

 

I am really confused on these things. I hope somebody can help me.

 

Regards

Klaus

 

 

1 ACCEPTED SOLUTION

Accepted Solutions

Re: Factor Constraints disappears center points and some other strange behavior

I understand you. What I am saying is that while you can imagine how it is possible to do it, Custom Design currently cannot do it. So you might have to find a way to manually add such runs. Perhaps fully manually, perhaps getting JMP to enter something as a useful starting point, and so on.

View solution in original post

6 REPLIES 6
statman
Super User

Re: Factor Constraints disappears center points and some other strange behavior

Just a couple of quick thoughts.  

1. Center points require all of the factors be continuous (or at least some scale where there is an actual center point), so if you have a categorical factor there is no possibility to get a true center point.  You can add three levels to some of the continuous variables if you'd like which will allow for estimation of a quadratic effect.

2. The way you describe the storage place, this appears to be nested in storage period, as all combinations are not permissible.

3. For the storage period where the value 2,4 occurs, this sounds like a software bug.  I could not reproduce the problem (I get values of 24 when I try it).  What version of JMP are you using?  What operating system?  You might report the issue to jmp support.

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

Re: Factor Constraints disappears center points and some other strange behavior

Hello,

thanks for your reply, Let me tell you my ideas to the problem and your replay:

I tried both, continous and categorial factors. In both cases JMP constructed center points for all the continous factors, not for the categorial ones. That was o.k. for me. Now, if in the same design I introduced the factor constraint, the column for center points disappears. That's what I do not understand. By the way, it is no problem to implement the center points manually.

 

I did not think, that there are nested factors (storage location, storage period). In my opinion nested would be the first picture, my design looks like the secon picture. May be I am wrong.

 

DHBWDozent_0-1626853381634.png

Regards

Klaus

Re: Factor Constraints disappears center points and some other strange behavior

It turns out that the Custom Design user interface appears to allow some combination of experimental features when, in fact, they are not allowed. Categorical factors introduce the problem of optimally constructing runs with continuous factors set to their mid-point but what of the categorical levels? Similarly, Discrete Numeric factors may get into a conflict with the given levels. Space Filling designs and hard-to-change factors generally do not work with center points.

 

Custom Design is mature and can handle many real world cases, but not all of them yet.

DHBWDozent
Level II

Re: Factor Constraints disappears center points and some other strange behavior

Thanks for your answer. But what I still do not understand is, that something is allowed/possible (the center points for the continous factors and only for them) if there are no factor constraints, and is disallowed if the only change is the implementation of a factor constraint, which has, in my opinion, nothing to do with the center points allowed without the constraint.

 

 

Re: Factor Constraints disappears center points and some other strange behavior

I understand you. What I am saying is that while you can imagine how it is possible to do it, Custom Design currently cannot do it. So you might have to find a way to manually add such runs. Perhaps fully manually, perhaps getting JMP to enter something as a useful starting point, and so on.

DHBWDozent
Level II

Re: Factor Constraints disappears center points and some other strange behavior

Hello,

yes, that is what finally works.

 

Regards

Klaus