時間の経過とともに観測値のいくつかが母集団を移動する場合の判別分析について

抄録

In this paper Fisher's linear discriminant function is investigated in which some data move from one to the other of two populations in the process of time. Utilizing the model for such kind of data, the mean vectors, common covariance matrix, discriminant coefficient vector, and cutting point are derived for each time. Two theorems for the coefficient vector and cutting point are obtained and certain numerical results for them are presented. From these theorems and numerical results, it is shown that if the data which change populations are close to the boundary of the end population, they do not greatly affect the coefficient vector and cutting point. Lastly, we show the example of prevention of geriatric diseases and discuss the coefficient vector and cutting point.