DOE Augment Design with Covariates

[hengehold_da1]

What inspired this wish list request?  Experimental units with known (fixed) covariates must be recruited and randomized to factor(s) over time so that they are independent (or "balanced") for the covariates.  Currently, there is not a automated way to do this in JMP 18.

What is the improvement you would like to see?  I would like to see JMP allow covariates in the the Augment Design platform so that new experimental units with known fixed covariates can be randomized to design factors that maintain balance of the existing design.

Why is this idea important?  This is important to design studies with important known covariates and staggered recruitment. 

[mia_stephens]

Thank you for submitting this suggestion. We are evaluating for consideration in a future release.

[MathStatChem]

It seems you can do this already, to some extent.  I would add to this request another idea:  partial covariates.  That is, I have a table with covariates, but some of the covariates do not have factor levels specified.  It would nice to be able to include them and have JMP fill in the missing factor levels using an optimal design approach.  An example would be I am planning a DOE, come up with an initial design that everyone agrees on, and then 1 or more additional factors are added to the study.  Some prespecified runs (e.g expected worst case combinations, centerpoints) are desired to include in the design, so those factor levels would be fixed for the new factors.  Rather than starting over, I would like to use the initial design, treating the initial design like covariates and then fill in the remaining missing factor levels on the new factors to give a optimal design.