How are odd factor settings in D-optimal RSM generated

[#1]

Dear JMP expert,

I set up two RSM designs, one with 3 factors (two continuous and one discrete numerical with four levels) and one with 4 factors (same as first plan but one more factor) using the custom designer (See Excel sheet with blinded factors). 

Now I have to explain my colleges in easy words, why I have odd factor settings (see red marked in excel sheet), which is difficult for me since I am not a hardcore mathematican and I just trust the JMP algorithm for custom design.

Thanks a lot for your help!

[#2 · Accepted solution]

Hi @DualARIMACougar,

The "strange" values obtained by the Custom Design platform are the results of the convergence of the coordinate-exchange algorithm used to compute optimal design.
See the response here : https://community.jmp.com/t5/Discussions/Random-decimals-incorporated-in-mixture-screening-design/m-...

You can still change these values and round up to the closest value (like 45 instead of 46,85 in your first design example), you may lose a little bit of optimality, but you can compare the initial design values with the new "corrected" values with the Compare Designs Platform to assess the influence of this/these change(s).

Hope this will help you,

[#3]

Thanks a lot Victor! This helps!