Opposite to augment DOE, hou could I quickly determine the DOE I need to remove to maximize the D-efficiency of resting DOE?

[#1]

Hi Pals,

I have collected 30 historical DOE which is not orthogonal at all. On JMP 17, the resulted D-efficiency is only 32 so I want to remove the most 1~3 nonorthogonal DOE before I augment any DOE. Is there any quick way to do that? Thanks! Currently, I can only manually recoginize the most nonorthogonal DOE from scatterplot matrix based on the design uniformity and re-calcuate the D-efficiency after remove it. 

[#2 · Accepted solution]

Hi @Fruit325,

One option could be to store all the runs in a data table and use them as candidate runs in the platform Custom Design to select the most informative runs with a D-Optimality criterion.
See https://www.jmp.com/support/help/en/17.2/#page/jmp/factors.shtml

Hope this option may help you,

[#3]

I agree with @Victor_G . That is how you should do it.

It depends on the nature of the data, of course, but removing 1 to 3 observations might not make a big difference to the D-efficiency. You may also need to think about augmenting with new runs to get a more useful data set.

[#4]

Hi @Victor_G 

Thanks! That's what I need. It works!

Hi @Phil_Kay 

You are right. Removing 10 runs still keeps a non-orthogonal design.... I will do augment DOE then. Thanks!