摘要
文章针对定时截尾试验的弊端提出了一个新的寿命试验方案,基于试验数据得到了似然函数,运用极大似然法得到了参数的迭代方程。利用EM算法寻求更优良的迭代方程,并根据缺损信息原则计算了Fisher信息矩阵。利用极大似然估计(MSE)的渐近正态性,推导出参数的渐近置信区间。运用Monte Carlo方法对估计的平均相对偏差(ARE)、均方误差(MSE)和平均区间长度(AIL)进行了模拟计算,并讨论了样本量对估计精度的影响。结果表明:定时区间删失样本所得到的相关估计量都要优于定时截尾样本。说明定时区间删失样本可以提高估计的精度。最后通过一个生存时间服从Rayleigh分布的例子,进行了相关统计推断。
This paper aims at the disadvantages of timed truncation test to propose a new life test scheme. The likelihood function is obtained based on the test data, and the iterative equation of parameters is obtained by using the maximum likelihood method. The EM algorithm is used to find better iterative equations, and the Fisher information matrix is calculated according to the defect information principle. The asymptotic normality of maximum likelihood estimation(MLE) is used to derive the asymptotic confidence interval of the parameters. The paper also uses Monte Carlo method to simulate and calculate the average relative error(ARE), mean square error(MSE) and average interval length(AIL) and discusses the influence of sample size on estimation accuracy. The results show that the correlation estimators obtained from the time-interval censored samples are better than those obtained from the time-truncated samples, indicating that the precision of estimation can be improved by deleting samples from time interval. Finally, the paper uses a case of survival time obeying Rayleigh distribution to make relevant statistical inference.
作者
龙兵
候兰宝
Long Bing;Hou Lanbao(School of Mathematics and Physics,Jingchu University of Technology,Jingmen Hubei 448000,China)
出处
《统计与决策》
CSSCI
北大核心
2020年第13期43-47,共5页
Statistics & Decision
基金
湖北省教育厅重点科研项目(D20184301)。