Subscribe Bookmark RSS Feed

Number of runs for a 4-factors DSD

dp30

Community Trekker

Joined:

Jun 23, 2011

Hi,

When using JMP11, the number of runs for a 4-factors DSD is N=9.

With JMP12, N=13.

Any reason for this?

Thanks

1 ACCEPTED SOLUTION

Accepted Solutions
Solution

I believe the number of recommended runs for this very powerful new screening design have evolved based upon experience. In general screening designs imply many factors and the power of the DSD's is experienced by leveraging the effect sparsity principle. With low numbers of factors there is a better chance that effect sparsity will not hold and you will need additional experiments to delineate the model. A better scenario would be to brainstorm an additional 2 factors since the number of experimental runs would be the same (13).

2 REPLIES
Solution

I believe the number of recommended runs for this very powerful new screening design have evolved based upon experience. In general screening designs imply many factors and the power of the DSD's is experienced by leveraging the effect sparsity principle. With low numbers of factors there is a better chance that effect sparsity will not hold and you will need additional experiments to delineate the model. A better scenario would be to brainstorm an additional 2 factors since the number of experimental runs would be the same (13).

ryan_lekivetz

Joined:

Nov 1, 2013

With the 9-run design, it is difficult to separate second-order effects. If you look in the comments of Why definitive screening designs are definitive, Bradley Jones discusses this in some detail. Due to this, we made the change in JMP 12 to give the larger design for 4 or fewer factors.