Resolution III Factorial Design for a Screening Experiment

Hi,

It will be greatly helpful if anybody responds to my question.
When I decided a resolution III factorial design, 2(5-2) for a screening experiment, a 8 run fractional factorial design by Minitab was generated and alias structure is given to me like below.
Design Generators: D = AB, E = AC

Alias Structure

I + ABD +ACE+BCDE
A + BD + CE + ABCDE
B + AD + CDE + ABCE
C + AE + BDE + ABCD
D + AB + BCE + ACDE
E + AC + BCD + ABDE
BC + DE + ABE + ACD
BE + CD + ABC + ADE

What I understand is main effect and two factor interactions should not be aliased if two factor interactions may affect the response during Resolution III design.

My concern is D+AB. For example, this screening experiment is associated with mold validation and if I assign A as temperature and B as time, I think that two factor interactions AB will possibly affect the response. Then we don?t know the effect will be the result of D, AB or both because main effects A and B are expected to affect the response most.

My questions is that it is an acceptable or right way to reassign main effects sequence (ex. assign temperature to main effect ?D? and time to main effect ?C?) to avoid for main effects to be confounded with two factor interactions which could affect the response.

Thank you very much for your help in advance.

Peter

A couple of questions about the purpose of your experiment. You mentioned that the experiment was for mold validation.

Are you just trying to see whether you can make acceptable product under the limits of mold process variation? If this is the case, you really don't need to worry about interactions. You are just using the DOE as an efficient means of testing the overall effect of extreme process conditions.

Are you using the design to identify which process variables are important and the optimum level at which to set them? Yes, you can take the approach that you mentioned about taking care which factors are assigned to which letters. However, unless you have theoretical or prior knowledge there will always be uncertainty.

There are a number of ways to reduce this uncertainty: 1) follow up with a fold-over design and analyze the combined design; 2) follow up with with a smaller design with the non-significant factors removed; 3) follow up with a confirmation run at the proposed new conditions and determine whether this matches the predicted results.

Dear Miner

I really appreciate for your advice. Yes, we are considering a fold-over design if there is uncertainty. I just want to confirm what I am trying to do is an acceptabe/right way if we are very confident that two factor interactions confounted with main effect will affect the result according to our experience.

Again, thank you very much for your help.

Peter