Is it possible to have 3 blocking variables in Custom design in JMP?

[#1]

Hi,

I was using JMP's custom design option to develop a design for a screening experiment. I am using custom design since I have mixture variables. I know it is possible to have 2 blocking variables in the design using the method mentioned here: https://community.jmp.com/t5/JMP-Blog/Constructing-a-custom-design-with-2-nested-random-blocking/ba-...

But is it possible to have 3 blocking variables in the design?

[#2]

Your link is broken so I could not review that particular blog post but I believe that it is the one by Bradley Jones in which he uses 'dummy' or 'fake' hard-to-change factors and very-hard-to-change factors to create whole plots and sub-plots. The latter are used to estimate the random effect of the blocking, assuming that the effects are not fixed. The former are deleted from the model (fixed effects) and the data table.

This approach can be extended to more than two blocking. The limitation is that JMP handles tri-randomization at most in Custom design. So you can work with whole plots, sub-plots, and within sub-plot error, but no more. The nesting requirement will prohibit using custom design with more than two nested bocking factors as far as I know.

[#3 · Accepted solution]

As Mark mentioned, there's currently no easy way to do this within Custom Design. It is high on our development wishlist for a future version.

If they don't need to be nested, you can try using fixed blocks (Add Factor->Blocking) and switch them to Random for analysis.

If you're looking at a split-split-split-plot, a possibility to create the design as described in the blog post, but set it up in such a way that you can break up the subplots by replacing the levels with 2 new columns that represent the blocking you want. This tends to work best if you're just looking to add random blocks however, otherwise you have to do a lot of manual manipulation to make sure that the factors obey the randomization that they should.

Cheers,

Ryan