Robust Parameter Design using Generalized Linear Model from a Bayesian Perspective

Robust parameter design has been performed extensively due to Taguchi's works. In most common cases, robust parameter design has been applied by assuming a linear model under normal distribution. Recently robust parameter design has been considered in generalized linear models (GLMs) for so many nonnormal problems in actual production.So far, some relevant researchers failed to consider noise variables' influences in dealing with nonnormal response problems while in most production the noise factors are the sources of the variation. In the article, a joint generalized linear model combining the mean and variance model is fitted from a Bayesian perspective with considering noise variables' influences. The feasibility of the model proposed is verified by a full factorial experiment with replicated measure.