Subscribe Bookmark RSS Feed

how to build an unbalanced split-plot DOE

johnsingh

Occasional Contributor

Joined:

Feb 9, 2017

Hello All,

 

I am building a DOE with experiments being carried over in two different labs (say, Lab A and B). In addition, there is a hard to change variable of temperature. So i decided to make labs as whole plots and temperature as sub plots. This gives me a DOE.

 

Now i have been told that one of the labs cannot really do all the experiments. In other words, the DOE i got had assigned 50% of experiments to each lab. Is it possible to tweak the design so that, say only 25% of experiments are performed in Lab A and the rest in Lab B.

 

Will appreciate any help on this.

John

1 ACCEPTED SOLUTION

Accepted Solutions
louv

Staff

Joined:

Jun 23, 2011

Solution

One approach to handle your scenario would be to build a JMP table that represents the number of experiments Lab A and Lab B can handle. Let's say for example 8 for A and 16 for B. Then add a column to that table to create a random number and then sort the table randomized. This table can then be brought into the custom design platform as a covariate and the Temperature made to be the hard to change factor to give your desired experiment to accomodate the Lab experiment constraint.

Screen Shot 2017-03-01 at 11.31.59 AM.png

4 REPLIES
louv

Staff

Joined:

Jun 23, 2011

John,

Have you considered doing all of your experiments in Lab A and then verifying the model from Lab A in Lab B?

johnsingh

Occasional Contributor

Joined:

Feb 9, 2017

We are trying to balance the workload between the two labs.

I would also prefer to get all experiments done in a single lab, but seems unlikely.

louv

Staff

Joined:

Jun 23, 2011

Solution

One approach to handle your scenario would be to build a JMP table that represents the number of experiments Lab A and Lab B can handle. Let's say for example 8 for A and 16 for B. Then add a column to that table to create a random number and then sort the table randomized. This table can then be brought into the custom design platform as a covariate and the Temperature made to be the hard to change factor to give your desired experiment to accomodate the Lab experiment constraint.

Screen Shot 2017-03-01 at 11.31.59 AM.png

johnsingh

Occasional Contributor

Joined:

Feb 9, 2017

Thank you for the response. It makes sense and i will give it a try.