摘要
Bayes逐步判别法是经济预测中常用的方法。该方法有两个基本假设:母体正态分布假设和各母体协方差矩阵基本相同假设,因此该方法在使用时有很大的局限性。本文从马氏距离的概念出发,以各母体先验概率为权,改变各统计样本的重要性,得出了在母体任意分布情况下,与Bayes逐步判别法相类似的表达式。并且证明了各母体协方差矩阵有较大不同时,判别结果的变化也是很小的。通过用本文方法对乌鲁木齐市、海城镇、营口市、唐山市及东川市的震害进行判别,证明是可行的。
Bayes seeeessive criterion is a method often adopted in economic prediction. In this study, we improve the limition existing in its hypothesis to derive the expression similar to that of Bayes seccessive criterion and apply our method to idenlify the earthquake disasters in Wulumuqi, Haicheng, Yingkou, Tangshan and Dongchuan, the results are convincible.
关键词
震害指数
先验概率
逐步回归
马氏距离
Index of earthquake damage
Prior probability
Successive recursion
Marhoff’s distance
