Using Logistic Regression to Evaluate New Lid Designs

New product designs from competing vendors must often be evaluated in order to make the best decision on which design to use in manufacturing. In most cases, the designs must be challenged over a range of environmental factors, process tolerances, and incoming material variations. At times, the performance characteristic may take on the form of a passing or failing response. Due to the dichotomous nature of a pass/fail outcome, ordinary regression techniques based on a normal distribution of error terms with constant variance are not appropriate. In the case of a pass/fail response, a natural distribution to consider is the binomial distribution. The authors have exploited the field of generalized linear models, specifically logistic regression, to select a new lid design for a beverage container based on the correct distributional assumptions.