Abstract
A method for group discrimination is examined which treats the observations of an unknown group (test group) as a set, while the usual discrimination method treats each observation separately. Effectiveness of the method is confirmed by simulation. The method is applied to differentiate rice cultivars and producing districts based on the brown rice grain contour shape. Statistically significant differences among cultivars and producing districts are observed in the shape of rice grain, however, the differences are small and distributions of parameters representing the shape of rice grain contour closely overlap each other. The precise discrimination was not possible in the usual discriminant analysis of each grain. The method of group discriminant analysis showed more effective results in discrimination of rice cultivars and producing district for the case of closely alike populations.
