Subscribe Bookmark RSS Feed

What is the specific method that JMP uses to maximize desirability in the Profiler

ashwinearl

Community Member

Joined:

Nov 12, 2011

Hi,

I am using JMP to do Response Surface methodology (RSM).  When using the response profiler and desirability functions, I will select the maximize desirability and JMP provides the recommended variable settings to maximize d. My question is what method does JMP use to actually solve this multi-objective optimization problem?  I can't find any documentation that states the general algorithm that is followed?  I need to say how, at least in general terms, JMP solves this. If it's propriatary, than is there a general way of referring to it?

thanks

1 ACCEPTED SOLUTION

Accepted Solutions
Solution

To maximize the desirability for continuous factors, JMP uses a constrained Newton's method approach with step shortening for the desirability functionality. A reference for this is:

Numerical Methods for Unconstrained Optimization and Nonlinear Equations by Dennis and Schnabel.


For categorical factors, JMP uses a greedy, one variable at a time optimization algorithm. When there are both categorical and continuous factors, JMP goes back and forth between the two strategies. To reduce the odds of finding a local optima, JMP uses multiple random starts. The optimizer chooses and alternates between a variety of optimization steps depending on the problem, and how it is behaving.

6 REPLIES
David_Burnham

Super User

Joined:

Jul 13, 2011

My guess would be that they use the Nelder-Mead Simplex Method.  Whilst the documentation does not explicitly state the method used (as far as I can see), the references include a number of papers relating to simplex optimisation.

Dave

-Dave
David_Burnham

Super User

Joined:

Jul 13, 2011

I'd quite like the know the official answer to this.  I get asked this question a lot.

-Dave
Solution

To maximize the desirability for continuous factors, JMP uses a constrained Newton's method approach with step shortening for the desirability functionality. A reference for this is:

Numerical Methods for Unconstrained Optimization and Nonlinear Equations by Dennis and Schnabel.


For categorical factors, JMP uses a greedy, one variable at a time optimization algorithm. When there are both categorical and continuous factors, JMP goes back and forth between the two strategies. To reduce the odds of finding a local optima, JMP uses multiple random starts. The optimizer chooses and alternates between a variety of optimization steps depending on the problem, and how it is behaving.

klinkert

Community Member

Joined:

Sep 7, 2015

Susan, I am using JMP for my dissertation, and I need to explain mathematically exactly how I arrive at my solutions (using JMP). I obtained the book you referenced, but did not find "step shortening" in this rather dense book.  In a very respected source of statistical methods, The NIST Engineering Statistics Handbook [2], I attempted to solve using JMP the example provided there. That example is by the originator of the desirability method (Derringer and Suich 1980 [1]).  Using JMP, I could not get the same result provided there.  My research problem has all continuous factors. Maybe you could please try it and see what you get. But most importantly, could you please provide a reference scientific paper for the method JMP uses for desirability, or at least the page numbers in the Dennis and Schnabel book that explains the mathematical method used to compute desirability for multiple factors. Thanks so much.

[1] Derringer, G., and Suich, R., (1980), "Simultaneous Optimization of Several Response Variables," Journal of Quality Technology, 12, 4, 214-219.

[2] 5.5.3.2.2. Multiple responses: The desirability approach

michael_jmp

Staff

Joined:

Jun 23, 2011

Hi Dave,

I've attached a draft of the updated doc for JMP13 regarding how the desirability optimization happens. I hope it helps address your question.

Thanks,
Michael

Michael Crotty
Sr Statistical Writer
JMP Development
David_Burnham

Super User

Joined:

Jul 13, 2011

Susan and Michael

Thank you both for the information that you provided.  Very helpful,  Regards

-Dave