In need of some help getting started with some experimental design work I am about to undertake. Due to the nature of my measured response (binary outcome - Defect (Y/N), it is my understanding that utilizing logistic regression to analyse the outco



In need of some help getting started with some experimental design work I am about to undertake. Due to the nature of my measured response (binary outcome - Defect (Y/N), it is my understanding that utilizing logistic regression to analyse the outcome is the way to go. My questions arise on testing design. I have constructed a 2^(7-3) fractional factorial testing my critical to quality variables. Based on your experiences, what would be an ideal sample size per 16 run opportunity. I realize that I am dealing with an odds ratio so the answer has a lot to do with the probability of finding the defect. That probability is realatively low under current state operating conditions but I really believe that I will capture the opportunity with my factor selection and settings. Will 10 parts per run, 160 total for the experiment going to be enough to speak to significance or would you suggest something else. The larger the sample the more accurate the model perhaps but I am dealing with time constraints