Analysis of a Definitive Screening Design Experiments

[#1]

Hi,

I've performed an 18 run 4 continuous factor definitive screening design experiment with center points.

I expected to be able to include the estimation of quadratic effects in my analysis, however, I'm only able to estimate them up to two way main effects and not quadratic which is supposed to be a capability of definitive screen designs. 

I followed the "fit screening design" option, I've included pdf copies showing what I've got when then making the models, one is up to factorial degree 2 (a pretty good analysis), and when I tried to include quadratic effect estimates aswell (which looks wrong). I'm sure (hope) there's something obvious I've missed (this is my first DSD) but any pointers would be appreciated. 

Thanks very much   

[#2 · Accepted solution]

For a DSD, main effects are uncorrelated with quadratic terms and 2 factor interactions (FI). However, quadratic effects are correlated with each other and 2 FIs, A DSD will also allow you to pick any 3 factors and estimate the full RSM. You have four factors in your model. Drop any fourth factor (main effect, quadratic, and any 2 FI that contains it) and you will be able to fit the model.

BTW, you could have used Custom Design to fit the full RSM with 4 factors. The precision might not be great, but all the terms would have been estimable.

[#3]

Hi Don, thank you very much, this is very helpful and just in time for the talk with my team about it