2003 年 11 巻 6 号 p. 58-63
When using the MTS method, we need tocal culate the inverse matrix in the course of calculating Mahalanobis distance. In such a case, we cannot avoid the problem of multi-colinearity. That is, if multi-colinearity exists, the rank of the matrix becomes lower and the inverse matrix cannot be calculated. To avoid this problem, some methods such as calculating the inverse matrix using a cofactor are proposed. Because the variable is not required for discriminating,the rank of a matrix might also become lower. We, therefore, discussed the problem of a matrix whose rank becomes lower when taking into account such multi-colinearity and unnecessary variables in the data.