536016_en-US

liliconcom

Plot of "normalized determinant" vs number of runs for Design Explorer

This is an idea from a chemometrics professor, Riccardo Leardi,  at University of Genoa. (Not a JMP user)

The "normalized determinant", M, is used as a measure of how good a design is.

The optimal design for different run sizes are determined. M is then plotted versus number of runs to suggest the appropriate number of runs to use.

 

This (or a similar metric) could be a useful plot for Compare Designs and/or Design Explorer.

There is more detail in this paper:

Leardi, R. (2022). D-Optimal Designs. In Encyclopedia of Analytical Chemistry, R.A. Meyers (Ed.). https://doi.org/10.1002/9780470027318.a9646

This approach can be utilized in CAT (Chemometric Agile Tool) software, which I think is based on R.  I have never come across anyone using this software, other than Prof. Leardi.

 

Design of ExperimentsRe: Plot of "normalized determinant" vs number of runs for Design Explorer

Thanks @SamGardner . Sounds good. I think relative D-efficiency makes a lot of sense. Although I suspect it will not be acceptable to Prof Leardi who is apparently convinced that there is only one "right" way to do optimal designs!

Re: Plot of "normalized determinant" vs number of runs for Design Explorer - Status changed to: Yes, Stay Tuned!Re: Plot of "normalized determinant" vs number of runs for Design Explorer

After some quick investigation, there are a number of suggestions provided for JMP EA17 that relate to this.  Updated to add the Jira issues.  Several additional visualizations are planned for design explorer, although we will probably use use relative D-efficiency vs using the D-efficiency or design matrix determinant.    There is a 1:1 relationship between the design matrix determinant and the D-efficiency.  

Re: Plot of "normalized determinant" vs number of runs for Design Explorer - Status changed to: InvestigatingRe: Plot of "normalized determinant" vs number of runs for Design Explorer - Status changed to: Acknowledged

Thanks @Phil_Kay for passing that idea along.  We will take this under consideration.