Subscribe Bookmark RSS Feed

Can a custom designed experiment have blocks of unequal size?

bwilliamson

Community Trekker

Joined:

Jul 27, 2015

I'm trying to put together a custom experiment that has the site of manufacture as a blocking variable. Each of the two sites has a different capacity for parallel work, making it desirable for me to have the matrix be unbalanced with respect to the number of conditions tested at each site so I can take advantage of the ~20% extra capacity at the larger site . Is it possible to set up a matrix that divides the experimental conditions between the blocks to my specification?

Thanks

1 ACCEPTED SOLUTION

Accepted Solutions
Solution

You will have to create a new data table with a column for Site with the number of runs equal to the number that you want, and the proportion of Site 1 and Site 2 set to the appropriate amount to account for the extra capacity.

In Custom Design, load the site variable as a Covariate. You can then add your other factors, and keep the run size the same when creating the design. You just have to be careful if adding interactions that you don't add any with the blocking variable, since the designer doesn't know it's a block. I'd also  recommend removing the interaction terms from the list of alias terms.

The loading of the covariate is similar to what I wrote about in this blog: Father's Day fun with toy cars and DOE

Cheers,

Ryan

2 REPLIES
Solution

You will have to create a new data table with a column for Site with the number of runs equal to the number that you want, and the proportion of Site 1 and Site 2 set to the appropriate amount to account for the extra capacity.

In Custom Design, load the site variable as a Covariate. You can then add your other factors, and keep the run size the same when creating the design. You just have to be careful if adding interactions that you don't add any with the blocking variable, since the designer doesn't know it's a block. I'd also  recommend removing the interaction terms from the list of alias terms.

The loading of the covariate is similar to what I wrote about in this blog: Father's Day fun with toy cars and DOE

Cheers,

Ryan

bwilliamson

Community Trekker

Joined:

Jul 27, 2015

Well, that was easy.

Thanks!