Abstract: Robust Design is an important method for improving product or manufacturing process design. In analyzing robust design experiments with a fixed target, Taguchi proposed a two-step procedure to identify the ``optimal'' factor settings that minimize the variance and adjust the mean to target. This paper generalizes the two-step procedure to a multi-step procedure based on the response model approach. We derive and illustrate the procedure under various conditions. We also show that the ``optimal'' solution from the procedure is invariant with respect to a general class of loss functions.
Key words and phrases: Taguchi's method, response model, loss model, product design, two-step procedure.