2014 Volume 22 Issue 2 Pages 30-37
By placing the unit space at the center of a distribution, the Taguchi-Schmidt (TS) method in the Mahalanobis-Taguchi system enables positive and negative predictions of target variables. With the Schmidt orthogonal expansion used in the TS method, however, the estimated results depend on the order of items,and the rules for ordering the items have been unclear, which has made the TS method hard to apply. Accordingly, the ordering of items in the TS method was examined through the use of financial data on the basis of two ideas; correlation coefficients and the combined estimation of the S/N ratio. The results suggest that it is appropriate to use the order that maximizes the estimated S/N ratio and an order that increases the correlation between items and target variables.
