Choose Language Hide Translation Bar
Highlighted
parag
Level I

D efficiency

I am new to JMP. In DOE, what is the significance of D efficiency and I efficiency. When I make a custom design model, should the D efficiency be under or over certain value for my design to be useful? In a model I made my D-efficiency is 74 % what does this mean? My design is good/bad/okay?

1 ACCEPTED SOLUTION

Accepted Solutions
Highlighted
louv
Staff (Retired)

Re: D efficiency

The measures of efficiency are scaled between 0 to 100

The efficiency is a measure of the goodness of a design relative to a hypothetical orthogonal design so they are not useful as absolute measures. They should be used in a relative fashion to compare one design to another. Efficiencies that are less than 100 may be perfectly satisfactory.

View solution in original post

5 REPLIES 5
Highlighted
louv
Staff (Retired)

Re: D efficiency

The measures of efficiency are scaled between 0 to 100

The efficiency is a measure of the goodness of a design relative to a hypothetical orthogonal design so they are not useful as absolute measures. They should be used in a relative fashion to compare one design to another. Efficiencies that are less than 100 may be perfectly satisfactory.

View solution in original post

Highlighted
louv
Staff (Retired)

Re: D efficiency

A great book on this topic is titled Optimal Design of Experiments by Peter Goos and Bradley Jones.

A key quote from the book with regard to this topic is

"do not recommend focusing on D-efficiency...we find it useful to compute relative D-efficiencies to compare competing designs"

Highlighted
parag
Level I

Re: D efficiency

Thanks a lot LouV. It helps me to feel better about my design!!!

Highlighted
Hamilto
Level I

Re: D efficiency

I have multiple designs with D, G and A Efficiencies of 100%. Since 100% is not possible are these good designs or should I reassess the number of factors and runs to obtain a more reasonable %? 

Highlighted

Re: D efficiency

100% efficiencies are possible. Your designs are likely just fine.

Dan Obermiller
Article Labels

    There are no labels assigned to this post.