Two step DoE

Question

Hello Commuinty,
perhaps it is a question already addressed, but could not find an answer yet. Specifically, it is about setting up a two-stage DoE. In my case a coating of two or more layers. Each layer has its own factors and there are therefore after each coating other responses that can be evaluated. Is there a clever way to create a DoE with JMP.

And even more specifically, do I have a way to add to a DoE for the coating of the first layer, after completing this experiment, other factors for a second layer to the existing DoE, taking into account the DoE's for the first layer? Thanks for the help!

statman · Answer

This indeed is a split-plot situation though it does not have anything to do with the difficulty of changing factor levels.  The first sequence of the experiment creates whole plot experimental units which can subsequently be treated with sub-plot factors to create the final experimental units for analysis.  I have not been able to find an automated way to create these types of designs (or the appropriate analysis) in JMP.  I have always done my own modified data table and analysis. (analysis requires separate tests for the whole plot and the sub-plot).

See:

Box, G.E.P., Stephen Jones (1992), “Split-plot designs for robust product experimentation”, Journal of Applied Statistics, Vol. 19, No. 1

MLanger · Answer

Thank you, that might help for new trials. But is there also a way to modify an already finished DoE with an additional stage and set of factors? To give an example. We finished a coating DoE that identified some levers but now there are some questions left for a not before considered second layer on top that comes with some new factors and we like to extend this experiment with the existing first time coated pieces.

Victor_G · Answer

Hi @MLanger,

Welcome in the Community ! :)

The situation you describe looks like a Two-way Split Plot experiment : a first step involving specific factors create batches, and then these batches are randomly applied with new factors from a second step process.

To create this specific design with the Custom design platform, the factors from the first step should be set as "Very hard to change", and the factors from the second step should be set as "Hard to change", with the option "Hard to change factors can vary independently of Very hard to change factors" checked (if you want to randomize applied factors from the step 2 in regards to step 1) :

Note that you can modify the assumed model, as well as the number of whole plot and subplots to your convenience and experimental budget.

You can find an example and explanation on how to setup and analyze this type of design here : https://community.jmp.com/t5/JMP-Sample-Data/Design-of-Experiments-Example-A-Two-Way-Split-Plot-Experiment/ta-p/22842?attachment-id=3233

If you need more details, help or guidance, you can provide more info on the number and type of factors, your goals and assumed model, etc...