576478_en-US

liliconcom

I Optimal design for mixture design

hi, we suggest to set the default criterium at I Optimal for a mixture design when using Custom Design

For the moment, using the custom design platform to create a mixture design, the default is a D Optimal design

For prediction purpose I Optimal will give a smaller integrated prediction variance

regards

François

 

Design of ExperimentsRe: I Optimal design for mixture design - Status changed to: Acknowledged

Hi @francois_berger, thank you for your suggestion! We have captured your request and will take it under consideration.

Re: I Optimal design for mixture design - Status changed to: InvestigatingRe: I Optimal design for mixture design

thanks for investigating and let me know if any question

NOTE: do you go to Discovery Europe?

Re: I Optimal design for mixture design - Status changed to: Not Planned For Now

Hi @francois_berger , it was nice to see you in Sitges!

We've discussed with the developer, and this would be a nontrivial change.

 

FYI - See this research paper comparing D- and I-Optimal mixture designs: https://www.tandfonline.com/doi/full/10.1080/01621459.2015.1136632 . 

Re: I Optimal design for mixture design

Just to add - the main issue comes when you have constraints, where you need to generate random points in the space. You can do that currently and it does generate I-optimal designs, but we require some further investigation for checking the consistency of the results. 

Re: I Optimal design for mixture design

thank you for the details and for the ref

François