Manually amending a DoE experiment suggested by JMP

[#1]

If for whatever reason, it is not possible to run a particular run, but a related one is possible, can the DoE design be manually amended? For example, if experiment ++-++ is suggested but is not possible and ++--+ is (and the latter does not appear elsewhere in the original design) can the user manually make the substitution, whilst bearing in mind that this change will likely alter the significance of the results generated?

[#2]

Yes. When you create a design in JMP...it's just a JMP data table that can be modified manually. The change may or may not alter the 'significance' whatever you mean by that word...but may change which terms are estimable in the model.

[#3]

Difficult to generalize with so little information. As Pete suggests, you can modify designs created by JMP, but this can impact resolution of the design (e.g., what terms can be estimated and which are confounded).  Why is that combination not possible?  Are your levels appropriate?  Should some terms be nested?

I'm not sure what you mean by "significance of results"?

[#4]

Dear @kjwx109prime , instead of amending a DOE afterwards, I think it would be better to give JMP that information of disallowed combinations upfront. Please see an example in the thread here: https://community.jmp.com/t5/Discussions/Disallowed-Combinations-not-working/m-p/780710#M96299 

BR

[#5]

Hi @kjwx109prime,

As @P_Bartell mentioned, the table generated by any DoE platform is "just" a datatable, you can always modify it afterwards.

However, as @statman mention it, the change of levels or factors settings in your DoE table can have some consequences on the performances of your design and the analysis done.

@Georg's advice is great, if you know the constraint beforehand, it's better tot take it into consideration during the design generation to ensure the generated design is optimal or near-optimal even with the added constraint.

I would recommend comparing the original and modified designs by using the platform Compare Designs. You can then have better details and informations about any possible lack of optimality, reduction in statistical power for detecting effects, increase of prediction variance or correlation between effects that the manual change of seting may have created compared to your original design.

You can then make a decision on how to continue : generating a new design with the constraint or continue with the manually modified design, thanks to the results of this platform.

Hope this helps,