ABSTRACT
Taguchi methods are widely used for parameter design to make products more robust to variations of environmental conditions. Taguchi proposed various performance measures known as Signal-to-Noise (SN) ratios for evaluating the performance of signal-response systems. However, the SN ratios do not always minimize quadratic loss.
In this paper, we describe a generalized SN ratio based on average log quadratic loss and K loss. This SN ratio is an equivalent of a performance measure for evaluating a variation related to the dispersion parameter for data with positive values.
Keywords: Loss function, Robust design, Signal-to-noise ratio, Taguchi method.