Weird level(s) assigned for a treatment when augmenting a screening DoE to an RSM

[#1]

Hi, 

We are augmenting a screening DoE to the RSM level and occasionally see a weird factor setting for one of the treatments. Should this value be used as is or can the value be changed to what its "supposed" to be? For example, the 0.032 level (highlighted in attached pdf file) was unexpected. Should this be used this particular experiment in the augmented DoE or can/should it be changed to 0.03?

We are using JMP 18.0.1.

Thank  you,

Mike

[#2 · Accepted solution]

Ok to change it to .03

[#3]

Thank you, Statman!

[#4 · Accepted solution]

Hi @mc1,

Welcome in the Community !

The Custom design (and Augment design) platform uses the Coordinate-Exchange Algorithm to create designs based on random points (and existing runs in the case of design augmentation), and then move the coordinates of the random points to improve the design based on defined optimality criterion and converge to a specific design. 

LinkedIn post on this topic : https://www.linkedin.com/feed/update/urn:li:activity:7163071856332234752/

The "stange" and very precise values you see are a result of this algorithm computation.

See more in about this already discussed topic here :

Random decimals incorporated in mixture screening design 

How are odd factor settings in D-optimal RSM generated 

Hope this answer will help you,

[#5]

Thank you, Victor!