Help for Some Sort of Nested Design

[#1]

Hello!

JMP newbie here.

I am working on a project where I need to come up with some sort of nested DOE design for some variables I want to test. I work in the proteomics industry and need to run an effects model of common med tech errors. We run our samples in batches of 96. There are some variables I need to test on the individual sample level (Like a reagent level differs for that sample), and others I need to test on the batch level (Like there are temperature fluctuations for this batch).

Does anyone know how I could set this kind of thing up in JMP?

Thanks for any help or resources!

[#2]

Hi @PowerJackal4255,

Welcome in the Community !

I think a Split-Plot design may be relevant for your needs. Split-Plot designs are designs where the randomization is restricted because of the presence of "Hard to change" factors. These factors are not randomly reset between each runs, but are changed from "whole plot" to "whole plot".

In your situation, factors tested on the individual sample level will be "Easy to change" factors, as their levels can be reset and changed randomly without restriction. Factors tested on the batch level will be "Hard to change" factors, as their levels are reset only at each new batch of 96 runs.

Here is how you can create this type of design with the Custom design platform (simple example here inspired from your use case):

1. Define your factors, and click on "Changes" next to each factor to change its randomization (depending if it's applied on batch (hard) or individual (easy) level) :
   

   Note that you can have several Hard to change factors. If you have several ones, you have the possibility to make them vary independantly from whole plot to whole plot, or not.

2. Then, you can add any constraints, and specify the assumed model you want to fit, as well as determining the number of whole plots (= number of batch in your example), as well as the total number of runs. In this example, since your samples are run in batch of 96, I defined 4 batches (whole plots) of 96 samples, so I defined 384 as my total number of runs :
    
3. Click "Make Design" to create your design. You can check that your design enable to have only one value for the hard to change factor for each whole plot :
   

You can read the documentation to see how to create and analyze such design here : Split-Plot Experiment

Some other ressources to help you:

• Personal LinkedIn post about Split-Plot design presentation and use case
• DoE Club session 2023 about Split-Plot designs : https://community.jmp.com/t5/Design-of-Experiments-Club/Recording-DOE-Club-with-Jacqueline-Asscher-a... 

Hope this answer will help you,

[#3]

Really hard to tell from your overly simple description of the situation.  In fact, it sounds like a nested sampling study vs. a DOE. If, however, you have factors that will be at different levels depending on other factors, for example, this is indeed a nested situation.  The key to handling nested variables is less in setup and more in analysis when you write the model.