How do I analyze a reaction work-up DOE?

[#1]

Hi all,

I am trying to analyze a work-up screening DOE. We did the experiments with two levels for 5 factors. The first three factors were grouped together as first variable and the two other factors grouped as second variable. With two center point runs, a total of six runs (00, -+, --, +-, ++, 00). How do I group the factors and fit a model to analyze this data in JMP16?

I attached a PGN file of the data I have.

Any help is greatly appreciated. Thank you. 

[#2 · Accepted solution]

Hi,

As you say, you have grouped factors 1-3 so that they are completely correlated. And you have grouped factors 4-5 so they are also completely correlated. Therefore, you only really have 2 factors that have been independently varied. It will not be possible to separately estimate the effects of factor 4 and factor 5, for example.

The best thing that you can do is to fit a model of each response vs factor 1 and factor 4. You will not be able to include factors 2, 3, and 5.

Use Analyze > Fit Model. Add your responses in the Y role. Ctrl-click to select factor 1 and factor 4, then Macros > Factorial to Degree. Run.

I hope this helps.

Phil

[#3]

Thank you for the response. 

[#4 · Accepted solution]

Welcome to the community.

In addition to Phil's comments, you could add a DF for the quadratic effect across the design space, so either Factor 1*Factor 1 or Factor 4*Factor 4.

The experiment is a full factorial on 2 factors with 2 center points.

[#5 · Accepted solution]

Did you consider any other designs? Do you need to isolate the effect of each factor in each group?

Custom design for the same main effects model would be larger but all the effects are separately estimable. Here is one candidate with 2 center points added.

A definitive screening design might be more appropriate for the same goal.

[#6]

Thank you.