对U检验法使用的探讨

An Examination on the Application of U-test

本文就假设检验中U检验法的使用不慎而对检验效果的影响通过数学推导逆行了分析探讨。指出在总体均值的假设检验中,当检验时采用的总体方差不是其真值时,必会导致检验的势函数值发生变化,检验中犯第一、二类错误的概率增大,同时对总体均值进行区间估计的置信度也受到影响。因此,左进行生产或工程实验中,为提高检验效果,少犯错误,应尽可能正确地推定新生产工艺或工程变更后的总体方差,不要盲目、简单地套用U检验法的检验程序。如果新总体的方差不易推定,则应采用T检验法进行检验。

This article analyses and studies, by mathematical inference,the effect on trist-ing results caueseci by careless use of U-test in the testing of hypothesis. It is pointed ous in the article that the value ol potential Junction in testing will necessarily change and the probability of making 1st and 2nd kind errors will be on the increase, and there is an effect on the confidence degree of forming interval estimation of population mean as well, when population variance taken by sampling is not equal to its real value in the process of testing of hypothesis on population mean. Therefore, in order to improve on testing results and reduce errors in production and engineering experiments, the poplation variance after altering new technology and engineering should be determined by deduction as correctly as it could be, and the routine of U-test cannot be applied mechanically in a simple and blind way. If it is difficult to determine by deduction the new population variance, then T-test could be applied to testing.