cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
JMP is taking Discovery online, April 16 and 18. Register today and join us for interactive sessions featuring popular presentation topics, networking, and discussions with the experts.
Choose Language Hide Translation Bar
ChrisLooi
Level III

Is Definitive Screening Design suitable for multiple responses?

Dear JMP Community,

 

In the case of DOE to screen factors for multiple responses, I try to use DFD (Definitive Screening Design).

 

The results however provided only screened significant factors for individual responses. I am more interested to know about significant factors affecting all the responses.

 

In the standard screening design (Fractional Factorial Res III, Res IV or Res V) for multiple responses, I can get results for significant factors that include all responses, rather than individually.

 

Is there anyway DFD can provide that kind of results for multiple responses?

 

Thanks and stay safe.

 

B.r,

Chris

1 ACCEPTED SOLUTION

Accepted Solutions

Re: Is Definitive Screening Design suitable for multiple responses?

The short answer is, yes, the definitive screening design is appropriate for the case with multiple responses.

 

The considerations for using a DSD over other design methods have nothing to do with the number of responses. They have more to do with deciding if this is a screening situation. That is, do you expect the key screening principles to hold and is it not one of the cases for which DSD was not intended (see Bradley Jones blog post).

 

If you are referring to using the Fit Definitive Screening option to select the model, then the responses are fit individually. This analysis scheme is not possible with more than one response at a time. You could save the fitted model for each response and then combine them by selecting Graph > Profiler. You would be able to examine how each factor affects all of the responses. (It is generally considered a best practice to fit the best model individually to each response.)

 

The other design choices produce the Model table script. JMP will jointly fit the multiple responses by default, but you can over-ride that choice. Fitting a common model still determines individual parameter estimates but the model contains the same terms for every response. The Effect Summary presents the result for each term for the response which is most significant (smallest p-value).

 

You can analyze a DSD experiment using the same Fit Least Squares platform by starting with Analyze > Fit Model if you prefer that behavior. You can always take control and choose the other behavior if it suits your purpose better.

View solution in original post

2 REPLIES 2

Re: Is Definitive Screening Design suitable for multiple responses?

The short answer is, yes, the definitive screening design is appropriate for the case with multiple responses.

 

The considerations for using a DSD over other design methods have nothing to do with the number of responses. They have more to do with deciding if this is a screening situation. That is, do you expect the key screening principles to hold and is it not one of the cases for which DSD was not intended (see Bradley Jones blog post).

 

If you are referring to using the Fit Definitive Screening option to select the model, then the responses are fit individually. This analysis scheme is not possible with more than one response at a time. You could save the fitted model for each response and then combine them by selecting Graph > Profiler. You would be able to examine how each factor affects all of the responses. (It is generally considered a best practice to fit the best model individually to each response.)

 

The other design choices produce the Model table script. JMP will jointly fit the multiple responses by default, but you can over-ride that choice. Fitting a common model still determines individual parameter estimates but the model contains the same terms for every response. The Effect Summary presents the result for each term for the response which is most significant (smallest p-value).

 

You can analyze a DSD experiment using the same Fit Least Squares platform by starting with Analyze > Fit Model if you prefer that behavior. You can always take control and choose the other behavior if it suits your purpose better.

P_Bartell
Level VIII

Re: Is Definitive Screening Design suitable for multiple responses?

I'm a little confused by the way in which you are using the phrases 'individual responses' and 'multiple responses'. If using the Fit Definitive Screening Design script (or the Fit Definitive Screening Design native platform) embedded in a JMP data table that was created using the Definitive Screening Design platform, I get separate and distinct analyses for each response in the report just as I would using the Fit Model platform. I don't necessarily get the SAME answers for a model, significant effects etc. because there is a built in logic/workflow for the Fit Definitive Screening script/platform in part because the workflow leverages some assumptions around effect heredity and sparsity. Details wrt to this workflow are in the JMP online documentation here:

 

https://www.jmp.com/support/help/en/15.1/#page/jmp/the-fit-definitive-screening-platform.shtml#