842082_en-US

liliconcom

DoE with Nested Factors

Hey all,

 

I have a customer asking about nesting factors in DOE - yes, there is a lot out there already, but I can't find an answer on his question.

 

He has

- 3 cell types/lines (A,B,C)

- 3 Donors for Cell type A

- 2 Donors for Cell type B

- 2 Donors for Cell type C

- some treatment factors.

 

So, there is some natural need for nesting the donors into the respective cell type. But this is where I struggle to understand what the approach should look like ideally.

I could simply create a design with a categorical factor "cell type" having 7 levels A1, A2, A3, B1, B2, C1, C2 and make it look like this:

 

But maybe someone here has a more appropriate way to set such a design up?

 

As always, appreciate your help and input

 

Best

Florian

Re: DoE with Nested Factors

You could include donor and cell type as seperate factors and make cell type a hard to change factor.  That will make it a split plot design where donors are randomly assigned within cell type.  When analyzing, be sure to include donor as a random effect.

Or are donors nested witin cell type because the specific donor has that specific cell type and therefore cannot be randomly assigned to cell type?

Re: DoE with Nested Factors

Hi @andreacoombs ,

 

thanks! Yes, the second case. They are investigating 3 different cell types, and have different numbers of donors  carrying each of these cell types.

Re: DoE with Nested Factors

How about treating Cell Type as a Covariate?

You'll have to create a table with one row for each cell donation type. The data doesn't have to be balanced, i.e., you don't need to have multiples of 3xA, 2xB, and 2xC although you can. Maybe one of the donors has contributed more material for the DOE. In the table I've attached, it is not balanced. The table doesn't have to be randomized, the Custom Designer will do that. Add a column corresponding to the Donor. 

Go to the Custom Designer. Add your 2 continuous responses. Select the Covariate table so it's the current table. Add a Covariate type factor and choose the covariate column. Build you're model and select the number of runs. Unless you check Allow covariate rows to be repeated, the Custom Designer will only use up to what's in the table. When you create the table, the row number corresponding to the covariate table is added. This way, you know which Donor to use for the cell type.

Re: DoE with Nested Factors

Thanks Don! Will propose this to the customer as well.