topic Re: Number of factor levels in I-Optimal design in Design of Experiments Club Discussions

Number of factor levels in I-Optimal design

SteveCz — Fri, 04 Apr 2025 16:59:58 GMT

For an RSM I-Optimal design created in Custom Design, is there a way to control the number of factor levels?

Here's an example.

I create an I-Optimal RSM design with 5 continuous factors and a few replicates and I get a plan with levels -1, 0 and 1 only
However, I have a keen interest in the behavior of one of the factors and thus prefer to have 5 different levels. A plan with -1, 0, 1 and two other levels (as determined by the design generator, not by me) would help me understand that factor better by filling in the space between -1 and 1 with 3 other levels, rather than only one.

Any sage advice would be much appreciated.

Steve

Re: Number of factor levels in I-Optimal design

Victor_G — Sat, 05 Apr 2025 07:01:59 GMT

Hi @SteveCz,

I see at least two options to get the desired number of levels in your DoE by using Custom design platform :

Either you specify in the model panel some higher order effects for your factors : for example, specifying up to 4th order polynomial terms in the model for X1 will create a design with 5 levels for X1,
Or you could create a design with discrete numeric factors in the Custom Design to "enforce" the number of levels and specific levels you want to include in your design, and switch back the design role of the factors to "continuous" once the datatable is created.

These are the options I already recommended in these discussions :

Hope you'll find a satisfying option for your needs,

Re: Number of factor levels in I-Optimal design

SteveCz — Sat, 05 Apr 2025 20:02:52 GMT

Hello Victor,

Thanks for the two suggestions. I tried both of the methods and they worked as you described without a major impact on power, prediction variance and the correlation coefficients from the Color Map. All up, a very good solution. Thank you!

Steve