【作者】 席同鑫； 林洪桦； 王中宇；

【Author】 Xi Tongxin~1 Lin Honghua~2 Wang Zhongyu~1 (1.School of Instrumentation Science and Opto-electronics Engineering Beihang University Beijing 100191,China; 2.Beijing Institute of Technology,Beijing 100089,China)

【机构】 北京航空航天大学仪器科学与光电工程学院； 北京理工大学；

【摘要】 在计量测试领域中,大多数随机误差按正态分布处理,正态分布是易于数据处理的对称分布。误差分布即便呈非正态性,仍大部分具有对称性。对称分布类误差的模型化及数据处理方法均较简便。在论文中论述了误差分布中对称性识别的必要性和可行性。将中位值与均值重合的识别、零偏度识别与秩和检验等方法应用于数据分布对称性识别,总结并归纳出不同方法的适用范围。通过蒙特卡洛模拟法得出测量次数为20和100的各组分布数值,利用三种方法分别求出不同的临界区间。通过测量实例验证了各种方法的正确性,推荐采用简便可靠的通用方法——中位值与均值重合识别法。利用matlab编程在微机上易于实现。更多还原

【Abstract】 In the metrology field most of the random errors were processed in the normal model,and the normal distribution was an easy data processing symmetrical distribution.Most of the errors were symmetrical even if they were not normal distribution.Symmetrical distribution of error data processing was simple and easy to model. In this paper the symmetry test methods for error distribution,their necessity and feasibility were discussed separately.The equivalence identification of mean and median,the zero skewness identification and the order sum test were all proposed.The application of different methods were summarized and concluded.20 and 100 of the various distributions numbers were got by the method of Monte Carlo simulation.The different critical intervals were calculated by using three methods.The results demonstrated the practicability of the new methods,and the simple and reliable method--the equivalence identification of mean and median could be recommended. Calculation program by using Matlab were made so that it could be easily realized on the computer.更多还原