【摘要】 本文采用的方法所建立的判别分析方程，适用于不满足正态性条件的因子数据．方法简单、 直观，且适应性强．本文仍选用熟知的逐步判别分析算例计算．结果为：本方法外延预报正确率 为１６／１７＝０．９４，而用逐步判别分析方法只达到１３／１７＝０．７６．更多还原

【Abstract】 In this paper we discuss how to establish the linear model y=β1X1+...+βmXm and let it possess stronger discriminatory ability to divide G classes of popu- lation A.(g=1. 2. "’ .G), when factor Xi (i =1 .2, .. ,m) possesses practical mea- nig with no model conditions of any kind. The discriminatory method of using modification power value in selecting factors is established. In order to test the model we compute an example, and the result is sa- tisfactory. For the sake of comparison we also use the stepwise discriminatory analysis tn compute the same example. and find that the accuracy of extellded forecast (16/17=0. 94)of our method is better than that (13/17=0. 76) computed by the latter one.更多还原