预测手段有效性的一个统计检验方法  

Statistical Test for Prediction Methods Efficiency

作　　者：刘嘉焜[1] 丁春蕾[1] 王公恕[2] 

机构地区：[1]天津大学理学院,天津300072 [2]南开大学数学科学学院,天津300071

出　　处：《天津大学学报（自然科学与工程技术版）》2004年第2期162-166,共5页Journal of Tianjin University：Science and Technology

基　　金：国家自然科学基金资助项目(重大90104015);南开大学 天津大学刘徽应用数学中心(T08)资助项目.

摘　　要：针对预报方法的"有效性"问题,以地震预报为例建立统计假设H0,即预测手段与地震流无关,通过构造适当的统计量,用统计检验的方法判别预测手段和方法是否有效.该方法判断结果错误的概率不超过预先给定值α=15%,从而为判断多种预报方法的有效性问题提供了一个检验方法.Different methods of forecasting natural disasters lead to different results.Aiming at the problem of forecast methods efficiency, this paper takes earthquake prediction as an example and makes statistical hypothesis H_0,in which forecast method is independent of earthquake fluid.Proper statistics variables were constructed and statistics test method was used to assess whether a prediction method is “efficient”.The probability of getting false result by this method was proved to be less than the given value α(α=15%).Thus a statistical approach to judging the efficiency of forecast methods was established.

关 键 词：预测手段 有效性 统计检验方法 地震 自然灾害 

分 类 号：P315.75[天文地球—地震学]