Practical Applications of Zero-Inflated Models in a Manufacturing SettingBrant Deppa, Winona State University
Mike Havener, Benchmark Electronics Inc
*Christopher J Malone, Winona State University
Keywords: Zero-Inflated Models, Manufacturing, Communication
A simple count is often used to quantify defects in a manufacturing setting. A manufacturing process of the utmost quality will have an over-abundance of zeros. As a result, commonly used statistically models (e.g. linear models) cannot be used to identify important factors in a designed experiment. Zero-inflated models account for the over-abundance of zeros and allow for the identification of significant factors that influence the presence, as well as the number, of defects. Fitting the actual models and identifying important factors turned out to be fairly straight-forward; however, communicating the results from these models to non-statisticians proved to be a challenge. A combination of tables and graphs were used to aid in the explanation of the findings. Specific examples of these will be presented.